Limited Effects of Variable-Retention Harvesting on Fungal Communities Decomposing Fine Roots in Coastal Temperate Rainforests

ABSTRACT

Fine root litter is the principal source of carbon stored in forest soils and a dominant source of carbon for fungal decomposers. Differences in decomposer capacity between fungal species may be important determinants of fine-root decomposition rates. Variable-retention harvesting (VRH) provides refuge for ectomycorrhizal fungi, but its influence on fine-root decomposers is unknown, as are the effects of functional shifts in these fungal communities on carbon cycling. We compared fungal communities decomposing fine roots (in litter bags) under VRH, clear-cut, and uncut stands at two sites (6 and 13 years postharvest) and two decay stages (43 days and 1 year after burial) in Douglas fir forests in coastal British Columbia, Canada. Fungal species and guilds were identified from decomposed fine roots using high-throughput sequencing. Variable retention had short-term effects on β-diversity; harvest treatment modified the fungal community composition at the 6-year-postharvest site, but not at the 13-year-postharvest site. Ericoid and ectomycorrhizal guilds were not more abundant under VRH, but stand age significantly structured species composition. Guild composition varied by decay stage, with ruderal species later replaced by saprotrophs and ectomycorrhizae. Ectomycorrhizal abundance on decomposing fine roots may partially explain why fine roots typically decompose more slowly than surface litter. Our results indicate that stand age structures fine-root decomposers but that decay stage is more important in structuring the fungal community than shifts caused by harvesting. The rapid postharvest recovery of fungal communities decomposing fine roots suggests resiliency within this community, at least in these young regenerating stands in coastal British Columbia.

IMPORTANCE Globally, fine roots are a dominant source of carbon in forest soils, yet the fungi that decompose this material and that drive the sequestration or respiration of this carbon remain largely uncharacterized. Fungi vary in their capacity to decompose plant litter, suggesting that fungal community composition is an important determinant of decomposition rates. Variable-retention harvesting is a forestry practice that modifies fungal communities by providing refuge for ectomycorrhizal fungi. We evaluated the effects of variable retention and clear-cut harvesting on fungal communities decomposing fine roots at two sites (6 and 13 years postharvest), at two decay stages (43 days and 1 year), and in uncut stands in temperate rainforests. Harvesting impacts on fungal community composition were detected only after 6 years after harvest. We suggest that fungal community composition may be an important factor that reduces fine-root decomposition rates relative to those of above-ground plant litter, which has important consequences for forest carbon cycling.

INTRODUCTION

Soil fungi are critical decomposers of plant biopolymers, such as those found in plant roots. Globally, fine-root litter accounts for nearly half of plant litter inputs into forest soils (1). Fine-root litter is a relatively recalcitrant material that accumulates as partially decomposed organic matter (2); up to 70% of the carbon stored in forest soils is of root origin (3, 4). Root litter represents a substantial energy source for fungi, and the composition of the root-decomposer community will likely influence ecosystem processes. This has been demonstrated in other substrates for decay rates (5) and carbon sequestration (6). Disturbance, such as forest harvesting, eliminates the flow of photosynthate used for growth by ectomycorrhizal fungi, dramatically shifting the below-ground fungal community toward saprotrophic dominance (7–9), potentially resulting in ecosystem-level increases in soil respiration and nitrogen mineralization due to ectomycorrhizal exclusion (10). Despite root litter being the dominant below-ground carbon source, the fungal community decomposing this substrate remains largely uncharacterized, and the response of these communities to forest harvesting is unknown.

The objective of variable-retention harvesting (VRH) is to retain a portion of the preharvest stand to preserve ecosystem functions. For below-ground fungal communities, VRH provides refuge trees that maintain populations of ectomycorrhizal fungi. Seedlings grown close to retention trees or patches have ectomycorrhizal communities similar to those found on retention tree roots (11–13). Retention patches may serve as a source of inocula for ectomycorrhizae in regenerating stands, though evidence indicates that their influence shortly after harvest does not extend beyond 10 m (14, 15). Kranabetter et al. found comparable levels of total ectomycorrhizal diversity in retention patches relative to intact forests 10 years after harvest (9). However, it is not clear whether VRH helps maintain ectomycorrhizal populations (via spore dispersal or direct root contact) through stand development and after a rotation.

Preservation of ectomycorrhizae via VRH may also influence carbon cycling, as these fungi are thought to play a role in organic matter decomposition. Saprotrophs have long been viewed as the primary decomposers of lignocellulose due to their extensive oxidative and hydrolytic enzymatic capabilities (16). In contrast, ectomycorrhizae are supported by host photosynthate, resulting in reduced genetic capacity for decomposition (17, 18). However, many ectomycorrhizae maintain some decomposer genes inherited from saprotrophic ancestors (19), such as those for the class II peroxidases involved in oxidative decomposition (20). Increasing evidence suggests that ectomycorrhizae use oxidative enzymes to mobilize nutrients complexed in soil organic matter (21). The net effect on decomposition is unknown and probably depends on the species present; although ectomycorrhizal species generally decompose less efficiently than saprotrophs (22), long-distance exploration types and species with greater oxidative enzymatic capacity (e.g., Cortinarius) may decompose organic matter more completely (20, 21). However, increasing evidence suggests that with access to host photosynthate, ectomycorrhizal fungi are better competitors for soil nitrogen, which leads to exclusion of saprotrophic fungi and subsequent decreases in soil carbon respiration (10). Thus, management practices that shift guild composition toward ectomycorrhizal fungi may reduce fine-root decomposition rates, at least shortly after harvest when saprotrophs would otherwise dominate under clear-cut (CC)harvesting.

The majority of fine-root litter inputs are in the organic and upper mineral soil horizons (23, 24), where ectomycorrhizal and other root-associated fungi dominate (25). Fungal succession during fine-root decomposition will likely be influenced by this surrounding fungal community, but this remains poorly studied. Cooke and Rayner describe three ecological strategies employed by fungi: R strategists are active in environments with high resource availability, S strategists are successful where a stress factor(s) excludes competitors, and C strategists succeed when stress or unexploited resources are reduced and competition is high (26). In this framework, succession during substrate decay typically follows a pattern where R strategists (e.g., molds and yeasts) rapidly colonize fresh substrates before competitive C strategists (saprotrophic and ectomycorrhizal basidiomycetes) invade and exploit the remaining resources. Indeed, this pattern is realized during leaf litter decomposition, where ascomycetes are often abundant early during decay and primarily carry out cellulose decomposition (27, 28). For fine-root decomposition, some evidence shows a high abundance of ascomycetous root endophytes (Phialocephala) and molds within the Mortierella genus after 6 months of decomposition, which were later replaced by basidiomycetes after 1 year (29). Stand age may also influence root-decomposing communities; in lower organic horizons in a boreal forest, saprotrophic fungi were more abundant in younger stands, ectomycorrhizae increased in abundance in older stands, and the ectomycorrhizal community composition shifted in older stands (30). Thus, both substrate decay stage and changes in overstory vegetation are likely to modify fungal communities decomposing fine roots.

Here, we compared fungal communities decomposing fine-root litter under VRH systems (aggregated and dispersed), in clear-cut, and in uncut stands in coastal Douglas fir forests on Vancouver Island, Canada (Fig. 1). Aggregate retention (AR) consists of small retention patches within a larger clear-cut matrix, whereas dispersed retention (DR) consists of uniformly distributed retention trees. We placed Douglas fir fine-root litter in these treatments replicated at two study sites, harvested in 2001 or 2008, and we sampled after 43 days and 1 year of decomposition. We hypothesized that VRH would result in a fine-root decomposer community with a greater abundance of ectomycorrhizal species and that this effect would persist 6 and 13 years after harvest (H1). We also expected to find a ruderal decomposer community in the early decay stage dominated by ascomycetes, which would be replaced by saprotrophic and mycorrhizal basidiomycetes after 1 year (H2). Finally, we anticipated that the fungal community would be structured by stand age, with increasing dominance of ectomycorrhizal guilds through stand development (H3).

FIG 1.

Schematic of sampling design (A) and root litterbag placement (B). Triangles represent sampling locations. Treatments are replicated at two sites, STEMS 1 and 3, harvested in 2001 and 2008, respectively. Sampling locations for clear-cut (CC) (n = 5) and uncut (UC) (n = 5) treatments were randomly selected. For the aggregate retention (AR) treatment, four retention patches were selected, and trenches were dug at the edge of the retention patch (AR-E, n = 4). For the dispersed retention treatment, four retention trees were randomly selected, and for each tree, a trench was dug 0.5 times the distance from the retention tree dripline to the center of the stem (DR-05, n = 4). Four additional “outside-area” sampling locations in each retention treatment were selected at least 10 m away from retention trees or patches to test for any spatial influence of VRH (AR-OA and DR-OA, n = 4 for both). At either site, the AR-OA, DR-OA, and CC samples were pooled for the stand age analysis to represent the 6- or 13-year-postharvest age classes (n = 4 + 4 + 5 = 13, for both). UC samples from both sites were pooled to represent the 70-year age class (n = 10). At each sampling location, a trench was dug (0.5 m long terminal repeat by 0.2 m wide by 0.2 m in height), and root litterbags were buried horizontally at the transition between LFH/Ae and B horizons or at 10-cm depth when no LFH existed (b).

RESULTS

Sequencing and community composition.

The numbers of reads that passed quality control were 67,276 of 183,659 and 93,083 of 213,191 for the 43 day and 1-year data sets, respectively. In the 1-year data set, the majority of reads were thrown out because of missing primers (18%) or missing tags (28%). Reads that were removed due to low quality or that were too short constituted 3 and 6% of total reads, respectively. A similar pattern was found in the 43-day data set. The mean read lengths were 295 and 270 bp for the 43-day and 1-year data sets, respectively. The average number of reads per sample was 1,326 ± 683, and clustering resulted in 649 operational taxonomic units (OTUs) (59 ± 18 OTUs per sample). Although some samples had low sequencing depth, Good's coverage, an estimate of sampling effort, was high, averaging 94.9% ± 0.02% across the 43-day and 1-year data sets. Coverage was significantly higher in the 1-year data set (93.2% ± 0.016% and 96.2% ± 0.015% for 43-day and 1-year data sets, respectively; F = 54.024, P < 0.001), likely due to the two additional single-molecule real-time sequencing (SMRT) cells used in sequencing the 1-year data set. All sequences are provided in Data Set S1 in the supplemental material.

Basidiomycota (59.8%) and Ascomycota (38.7%) were the dominant phyla, distantly followed by Zygomycota (1.2%), and Rozellomycota, Glomeromycota, and Chytridiomycota (all < 0.1%). Of the 649 OTUs, 86.3% and 32.4% were identified to genus and species, respectively. The top 112 and 43 most abundant species and genera represented 93.9% and 94.9% of the cumulative relative abundance, respectively. Genera with >2% abundance included Mycena (17.7%), Rhizopogon (9.1%), Phoma (5.6%), Hymenoscyphus (4.4%), Truncatella (4.2%), Serpula (2.8%), Meliniomyces (2.7%), Agrocybe (2.4%), and Phlebiella (2.1%). Similarly, Mycena leptocephala (13.6%), Rhizopogon vinicolor (8.3%), Phoma violacea (5.0%), Hymenoscyphus spp. (4.4%), Truncatella angustata (4.2%), Helotiales spp. (3.0%), Serpula himantiodes (2.8%), Mycenaceae spp. (2.6%), Meliniomyces variabilis (2.5%), Agrocybe praecox (2.4%), and Venturiales spp. (2.4%) constituted the top species with >2% abundance across the entire data set.

In the 43-day decay stage, plant pathogens were the dominant guild (31.5%), followed by saprotrophs (16.5%), ericoid mycorrhizae (10.4%), yeasts (7.9%), root endophytes (5.3%), ectomycorrhizae (4.9%), and molds (2.8%). After 1 year of decay, saprotrophs were most abundant (52.5%), followed by ectomycorrhizae (20.2%), plant pathogens (3.5%), ericoid mycorrhizae (1.4%), root endophytes (1.1%), yeasts (0.6%), and molds (0.3%). The proportion of fungi with unknown function was similar between decay stages (20.7 and 20.4% for the 43-day and 1-year decay stages, respectively).

Effects of harvest treatment and decay stage on fungal β-diversity.

The effects of harvest treatment and decay stage on fungal β-diversity (H1 and H2) were assessed together in a global permutational multivariate analysis of variance (PERMANOVA) model that included harvest treatment (aggregate retention [AR], dispersed retention [DR], clear-cut [CC], and uncut [UC]), decay stage (43 days and 1 year), site (13 years postharvest [STEMS 1] and 6 years postharvest [STEMS 3]), and their interactions (Table 1). No treatment effects were found in the model, but a significant decay stage, site, and treatment × site interaction was found (Table 1). This indicated that treatment effects differed depending on site, and therefore, each site was analyzed separately. Analysis of multivariate dispersion suggested group homogeneity for all main effects except time (see supplemental text and Table S1 in the supplemental material).

TABLE 1.

Effects of harvest treatment, decay stage, and site on fungal β-diversity assessed by PERMANOVA^a

PERMANOVA model and parameter	df	Sum of squares	Pseudo-F	Pseudo-r2	Pb
Global
Treatment	3	0.807	1.222	0.048	0.127
Decay stage	1	3.4402	15.631	0.204	0.001
Site	1	0.4633	2.105	0.027	0.015
Treatment × decay stage	3	0.5738	0.869	0.034	0.743
Treatment × site	3	0.9526	1.443	0.056	0.032
Sampling time × site	1	0.1668	0.758	0.01	0.774
Treatment × decay stage × site	3	0.5767	0.873	0.034	0.716
Residuals	45	16.885
Total	60
STEMS 1 (13 yr postharvest)
Treatment	3	0.7962	1.2488	0.1012	0.14
Decay stage	1	2.0484	9.6384	0.2604	0.001
Treatment × decay stage	3	0.5601	0.8785	0.0712	0.664
Residuals	21	4.463		0.5673
Total	28	7.8677
STEMS 3 (6 yr postharvest)
Treatment	3	0.9803	1.4412	0.1146	0.019
Decay stage	1	1.5371	6.7796	0.1797	0.001
Treatment × decay stage	3	0.5937	0.8729	0.0694	0.751
Residuals	24	5.4412		0.6362
Total	31	8.5523		1

^aHarvest treatments were uncut, clear-cut, aggregate-edge, and dispersed-0.5; decay stages were 43 days and 1 year; and sites were STEMS 1 and 3).

^bSignificant effects (P < 0.05) are shown in bold.

Decay stage was the most significant (P < 0.001) driver of community composition, explaining 20.4% of the variation in the fungal community across both sites and 26 and 18% of the variation at STEMS 1 and 3, respectively (Table 1; see supplemental text and Fig. S1 in the supplemental material). Treatment effects were present only at STEMS 3 (6 years postharvest), explaining 11.5% of the variation in the fungal community (Table 1) (P = 0.019). In the distance-based redundancy analysis (db-RDA) ordination for STEMS 3 (Fig. 2A), samples in the AR treatment and the UC treatment clustered together, and were separated from the DR and CC samples, which clustered together, but pairwise treatment contrasts showed no differences (see supplemental text and Table S2 in the supplemental material).

FIG 2.

Distance-based redundancy analysis (db-RDA) sample (A) and species (B) ordinations based on fourth-root-transformed fungal abundance at STEMS 3 (6 years postharvest). Only variation explained by decay stage (axis 1) and harvest treatment (axis 2) is visualized, with the percent variation explained in parentheses. For the species plot (B), only the top 50 most abundant species are visualized; species are color coded by functional group and are sized according to their relative abundance. The species epithet (when known) was removed to improve readability. Red shapes represent treatment centroids (aggregate edge, ■; clear-cut, ●; dispersed-0.5, ▲; uncut, ⬥). Time centroids are represented by hollow shapes (43 days, ○; 1 year, △). Dashed lines represent 95% confidence ellipses (calculated in vegan) around treatment or time centroids.

A similar analysis approach was used in the multivariate generalized linear model (GLM) analysis; in the global model, no treatment effects were detected, but site, decay stage, and all interactions were significant (P < 0.05) (see supplemental text and Table S3a in the supplemental material). As in the PERMANOVA analysis, the significant treatment × site interaction motivated separate analysis for each site. In general, multivariate GLM results were similar to those of the PERMANOVA analysis (see supplemental text and Table S3b and c); however, there was a significant treatment/decay stage interaction at both sites (see supplemental text and Table S3b and c), and all treatment contrasts for STEMS 3 (assessed via summary.manyglm) were significantly different (P < 0.05) (see supplemental text and Table S3d).

Responsive species and guilds.

At both sites, the effect of decay stage (H2) on species was dramatic; after 43 days, molds, yeasts, root endophytes, and especially ericoid mycorrhizae and plant pathogens constituted 62.3% of the mean relative abundance, but these guilds were reduced to 9.4% abundance after 1 year of decomposition and were replaced by saprotrophs (59.8%) and ectomycorrhizae (22.4%). All guilds significantly shifted (P < 0.05) between decay stages (Fig. 3A), as did 51 of the 163 most abundant species (P < 0.001) after adjustment for multiple testing (Fig. 4). Highly abundant species driving shifts after 1 year of decomposition included Mycena species and Hymenoscyphus spp. (saprotrophs) and Rhizopogon vinicolor and Tomentella sp. (ectomycorrhizae). Abundant fungi in the 43-day samples included Rhizopogon vesiculosus (ectomycorrhizal), Phoma herbarum and Trichoderma koningii (plant pathogens/saprotrophs), Meliniomyces variabilis (ericoid), Helotiales spp. (unknown), Cadophora luteo-olivacea (root endophyte), and many yeast and mold species (Fig. 4; see supplemental text and Fig. S1). The species plot of the global db-RDA model (see supplemental text and Fig. S1) was in general agreement with the species identified in the GLM analyses. Shannon diversity and evenness were also significantly higher (P < 0.001) in the early (43-day) versus later (1-year) decay stage (see supplemental text and Table S6 in the supplemental material). Basidiomycota and Glomeromycota were significantly more abundant after 1 year of decay, whereas Ascomycota were less abundant (P < 0.05) (Fig. 3B).

FIG 3.

Mean relative abundances of fungal functional groups (A) and phyla (B) in fine-root litter samples decomposed for 43 days or 1 year across all sampling locations. Only the globally most abundant species were functionally classified (90% of the reads) for panel A, and all reads were used for panel B. All guilds and Ascomycota, Basidiomycota, and Glomeromycota shifted significantly between decay stages (P < 0.05).

FIG 4.

Z-scores of fungal species abundance at the 1-year decay stage. Only species significantly responding to decay stage (43 days or 1 year) in a multivariate GLM model are visualized (P < 0.05, after adjustment for multiple testing using a step-down resampling procedure in the mvabund package). Positive values indicate a higher abundance of a given species at the 1-year decay stage, and negative values indicate a higher abundance at the 43-day decay stage.

The identification individual species and guild response to treatment effects at the 6-year-postharvest site (STEMS 3) was hampered by high variability between samples and a weak overall treatment effect, resulting in no treatment effects in the guild data (P > 0.05) (data not shown; see supplemental text and Fig. S2 in the supplemental material) and weak patterns in the species data. The two approaches used for identification of individual species response to treatment were also inconsistent. The species plot for STEMS 3 (Fig. 2B) identified ectomycorrhizae (Russula cessans, Tomentella spp., and Inocybe chondroderma), plant pathogens (Fusarium and Trichoderma), and saprotrophs (Mycena leptocephala and Phlebiella christiansenii) that appear to cluster within the overlapping data clouds for the AR and UC samples. Species clustering in the overlapping DR and CC data clouds include saprotrophs (Phlebiella spp., Plectania spp., and Mycena species), ectomycorrhizae (Rhizopogon vinicolor and Laccaria bicolor), and plant pathogens (Truncatella angustata and Pilidium concavum). Of the 77 species included in the STEMS 3 GLM, none varied in response to treatment after adjustment for multiple testing (P > 0.05).

Stand age.

Samples were grouped into age classes of 6, 13, and 70 years postharvest to assess the effect of stand age. The 6-year age class consisted of STEMS 3 CC, AR-OA, and DR-OA samples, the 13-year age class consisted of STEMS 1 CC, AR-OA, and DR-OA samples, and the 70-year age class consisted of STEMS 1 and 3 UC samples (see Materials and Methods for a description of AR-OA and DR-OA samples). When samples were grouped in this way, samples from the 13- and 70-year-old stands grouped together but were apart from the samples in the 6-year-old stands (Fig. 5A). Although samples from the 70-year stands were collected at two sites, this appeared not to influence community composition, as β-diversity did not differ between sites for these samples (P = 0.118). PERMANOVA and multivariate GLM models indicated a significant (P < 0.01) effect of stand age (see supplemental text and Table S7a and b in the supplemental material). An interaction was detected in the multivariate GLM models, indicating that effects of stand age differed depending on decay stage. Post hoc contrasts of the PERMANOVA results showed no difference in fungal community composition between 13- and 70-year-old stands, but the 6-year-old stands were significantly different from the 13- and 70-year-old stands (see supplemental text and Table S7c). Contrasts assessed in the multivariate GLM indicated unique fungal communities in each stand age (see supplemental text and Table S7d).

FIG 5.

Distance-based redundancy analysis (db-RDA) sample (A) and species (B) ordinations based on fourth-root-transformed fungal abundance. Only variation explained by decay stage (axis 1) and stand age (axis 2) is visualized, with the percent variation explained in parentheses. For the species plot (B), only the top 50 most abundant species are visualized; species are color coded by functional group and are sized according to their relative abundance. The species epithet (when known) was removed to improve readability. Red shapes represent stand age centroids (6 years, ●; 13 years, ■; 70 years, ◆). Time centroids are represented by hollow shapes (43 days, ○; 1 year, △). Dashed lines represent 95% confidence ellipses (calculated in vegan) around treatment or time centroids.

Ectomycorrhizae shifted markedly between the 6-, 13-, and 70-year-old stands, with each stand age maintaining a unique ectomycorrhizal community (Fig. 5A and 6). Ectomycorrhizae were less abundant in 6-year-old stands, except for R. vinicolor. In the 13-year-old stands, Amphinema spp., Wilcoxina rehmii, and to a lesser extent Tomentalla spp. and R. vinicolor were abundant. Tomentella and Pseudotomentella species, as well as Inocybe chondroderma, became more abundant in the 70-year-old stands (Fig. 5B and 6). Saprotrophic fungi were more abundant in the 6- and 13-year-old stands than in the 70-year-old stands, and the species composition differed between the 6- and 13-year-old stands (Fig. 6). One saprotrophic fungus, Gymnopus inusitatus, was abundant in the 70-year-old stands. Additional species abundant in the 13- and 70-year-old stands, but not in the 6-year-old stands, were identified from the species plot (Fig. 5B) and included the ectomycorrhizae Piloderma spp. and Russula spp. and the saprotrophs Phlebiella christiansenii and Resinicium bicolor. Mean guild relative abundance did not differ between stand ages (see supplemental text and Fig. S3 in the supplemental material), nor did alpha diversity or evenness.

FIG 6.

Z-scores of fungal species abundance in 6-, 13-, and 70-year-old stands. Only species significantly responding to stand age in a multivariate GLM model are visualized (P < 0.05, after adjustment for multiple testing using a step-down resampling procedure in the mvabund package). Species are organized according to functional group (ECM, ectomycorrhizal; SAP, saprotroph; RE, root endophyte; PP, plant pathogen; UNK, unknown).

DISCUSSION

We found similar species compositions in the 13- and 70-year-old Douglas fir stands, indicating that VRH has only marginal short-term effects on β-diversity in the fine-root decomposer community. Indeed, we found no evidence to support the hypothesis that ectomycorrhizal fungi were more abundant under either AR or DR or even in UC stands. Rather, we found that stand age significantly structured the species composition, again with no effect on guild composition. Decay stage was the only factor that resulted in significant shifts in guild composition. Taken together, our results indicate that successional patterns commonly seen in fungal communities through stand development are reflected in the fine-root decomposer community but that decay stage is more important in structuring fungal guild composition than are changes in the surrounding soil fungal community caused by VRH.

In partial agreement with our first hypothesis, we found that harvest treatment significantly altered fungal β-diversity at the 6-year-postharvest site, where species composition was similar between the UC and AR treatments and between the CC and DR treatments. Ectomycorrhizal fungi, being sensitive to harvest, appeared to be responsible for this. There was some indication that ectomycorrhizal species commonly associated with early and even late stages of stand development (Russula, Tomentella, and Inocybe) were more common under AR and in UC stands, whereas the disturbance-adapted fungi R. vinicolor and L. bicolor were more common under DR and in CC sites. Isolation of ectomycorrhizal hosts in DR may be responsible for this trend; ectomycorrhizal richness declines as host areal extent is reduced (9, 31). Aggregate retention patches may provide better refuge for fungi that are poor dispersers and are, therefore, more sensitive to host isolation caused by DR. Although aggregate patches cover a smaller area, it appears that dispersed trees do not maintain the conditions necessary for persistence of some ectomycorrhizal fungi early in stand development. However, our univariate GLM analysis did not identify any species responding to treatment, and no effect of harvesting was found after 13 years. Likewise, Varenius et al. found that management history (clear-cut versus shelterwood) did not result in differences in ectomycorrhizal communities in 50-year-old Pinus sylvestris stands (32). As well, contrary to our first hypothesis, harvesting did not result in a shift in guild composition, and there was no indication that VRH treatments resulted in greater dominance of ectomycorrhizal fungi. Reductions in root-associated (ectomycorrhizal/ericoid) guilds across LFH horizons that persist 12 years after clear-cut harvesting have been reported for Swedish P. sylversteris stands (30), and similar persistent trends 10 to 15 years after harvest were found in mineral and organic horizons in Pinus contorta stands in interior British Columbia (8). Our divergent results may reflect the high productivity in our temperate rainforests; the site index (50 years) is typically between 34 and 36 m for P. menzieseii at our sites, compared to 18 to 24 m for P. contorta in the study by Hartmann et al. (8). Relatively rapid regeneration may facilitate rapid reestablishment of ectomycorrhizal fungi in these stands within 6 years, resulting in the relatively stable abundance of this functional group on decomposing fine roots between harvested and unharvested stands.

In support of our second hypothesis, we found distinct fungal communities on roots sampled after 43 days versus 1 year of decomposition, where Ascomycota initially dominated but were largely replaced by Basidiomycota after 1 year. This is in agreement with fungal succession in leaf litter (27) and in decomposing woody debris (33). Consistent with fungal successional theory, we found that the fungal community at 43 days consisted of a diverse community of ruderal plant pathogens, root endophytes, molds, and yeasts. The plant pathogens Phoma herbarum and Fusarium spp. proliferated early in decomposition, likely in response to initial disturbance but also due to the availability of a labile substrate. We also found that ericoid mycorrhizae in the Rhizoscyphus ericae aggregate (including Meliniomyces spp.) initially proliferated but declined in abundance after 1 year. These fungi are considered stress tolerant and are often abundant under low-pH conditions favored by their Ericaceous host plants (34), but we found ruderal-like properties in these fungi. Indeed, the R. ericae aggregate was recently shown to possess a decomposer genetic profile similar to that of plant pathogens or saprotrophic generalists (35). The Ericaceous shrub Gaultheria shallon was highly abundant at all sampling locations, and disturbance of this root system during sampling may have severed ericoid mycorrhizal access to simple carbohydrates and forced upregulation of their decomposer genes, resulting in their proliferation at the earlier sampling time.

Most of the R strategists were transient, replaced after 1 year by a community of lower diversity and evenness characterized by the invasion of competitive C strategists such as saprotrophs in the genera Mycena, Phlebiella, and Agrocybe. These saprotrophs used fine roots as a carbon source, even though root litter was placed in lower horizons where ectomycorrhizal fungi are abundant. Ectomycorrhizae such as R. vinicolor, Tomentella spp., and Amphinema spp. were also abundant, with nearly a quarter of the reads assigned to ectomycorrhizae after 1 year. Li et al. found a similar guild composition in decomposing Pinus resinosa fine roots (29). This pattern of decomposition is very different from that for surface litter, where vertical stratification of the fungal community results in few root-associated taxa at a comparable stage of decay (22, 27). Fine roots typically follow asymptotic decay, where mass loss is initially rapid but is followed by a period of limited additional decay, whereas surface litter decomposes more completely before reaching an asymptote (36). We provide evidence that the period of initial rapid decay during fine-root decomposition is associated with molds, yeasts, and plant pathogens, a fast-growing community that is likely utilizing soluble sugars and starches that are more readily available early in fine-root decomposition (37). Although differences in substrate chemistry may partially explain the different patterns of mass loss between fine roots and leaf litter during later stages of decomposition, we speculate that asymptotic decomposition in fine roots may be related to the abundance of ectomycorrhizae in this substrate. Fine roots represent an overlapping niche for ectomycorrhizae and saprotrophs where competition is likely to be intense; saprotrophs compete for C, and ectomycorrhizae scavenge for complexed nitrogen (38). The result of this competition may retard root decomposition, an effect originally proposed for above-ground litter as the “Gadgil effect” (39). Due to difficulties in identifying field-collected fine-root species and in distinguishing between recently senesced and living fine roots, we used seedling (nonmycorrhizal) fine roots in place of those collected in situ. Although a caveat of this work, it is likely that ectomycorrhizal fungi would have previously colonized recently senesced fine roots, potentially intensifying competition between saprotrophs and ectomycorrhizal fungi and further retarding fine-root decomposition. We stress that if interguild competition plays a role in suppression of fine-root decomposition rates, this would need confirmation by directly linking decomposition rates with fungal community composition at the level of the individual sample.

In partial support of hypothesis three, we found clear evidence that stand age structured fungal communities decomposing fine roots, but there was no evidence that the proportion of ectomycorrhizal fungi increased over time. Differences in fungal communities between stand ages were determined largely by a succession of ectomycorrhizae. We found that the disturbance-adapted ectomycorrhiza R. vinicolor was most abundant in the youngest stands and was less abundant in older stands, consistent with the findings by Twieg et al. (40). We also found that the abundance of an Amphinema sp. and W. rehmii was higher in 13-year-old stands. These species have been identified as early-stage colonizers that may not be able to compete within a few years of regeneration (40–42). Kyaschenko et al. observed an increased abundance of species within the Atheliaceae (including Amphinema spp.) after 10 years of harvesting and suggested that this could be a result of adaptation to high N availability in younger stands (30). Indeed, Amphinema spp. and Wilcoxina spp. have been found in greater abundance with increasing site fertility (43). However, soil nitrogen concentrations initially increased in response to clearcutting but became negligible after 5 years in a forest type similar to that in our study (44). Host specificity is an unlikely explanation for the delayed establishment of these species, as both are known generalists (45), nor is dispersal limitation, as at least Wilcoxina is a widespread genus in North American spore banks (46), and Amphinema species have been reported in seedling bioassays within 5 years of harvesting (47). R. vinicolor may simply be more competitive early in succession but declined in response to increased competition from Amphenima spp. and W. rehmii. We also cannot exclude that these patterns of succession may be due to differences in moisture and nutrient regime between sites.

The 70-year-old stands were composed of a distinct ectomycorrhizal community, with tomentelloid (Tomentella and Pseudotomentella), Russula, and Inocybe species driving community composition. Although we found that tomentelloid and Inocybe species were abundant in our mature stands, they have been reported across a range of ecosystem development stages, including primary successional habitats (48, 49), recently cut/young stands (50–52), and mature stands (41, 53). Species in the Russula genus are more frequently encountered in mature forests (54), but have also been found fruiting in young clear-cuts (55). Generalizations about factors driving ectomycorrhizal succession through stand development are thus difficult to make, and the mechanisms behind these changes remain unclear, but they could be related to dispersal ability (56), priority effects (57), or edaphic conditions (58).

We also found evidence of shifts among saprotrophs with stand age that may be related to patterns of substrate recruitment during regeneration. The wood decomposers Agrocybe praecox and Pholiota spumosa were both abundant in the 6-year-old stands. These species may have responded to the abundance of logging residue shortly after harvest. As well, litter-decomposing taxa (Mycena spp.) were more abundant in the 13-year-old than in the 6-year-old stands, possibly due to increased litter inputs in the 13-year-old stands following stand regeneration. These substrate dynamics may explain why wood decomposers were less abundant in 13-year-old stands, possibly limited by short-term exhaustion of woody substrates and by competition from litter decomposers, which may become abundant as a result of increased litterfall. Similar successional patterns among saprotrophs in relation to substrate dynamics have also been reported in response to widespread insect disturbance; litter-decomposing genera peaked after bark-beetle-induced litterfall, followed by an increased abundance of wood decomposers associated with recruitment of deadwood (59).

Conclusion.

The recovery of the fungal fine-root decomposer community within 13 years of harvest, including in clear-cut treatments, suggests that the fungal community in this productive temperate rainforest is resilient to harvesting. Aggregate retention plots retained a fungal community similar to that in mature forests 6 years after harvest, suggesting that this method of variable retention moderates the effects of harvesting. Variable retention may be used to accelerate the development of structurally complex stands across landscapes (60), but it remains to be seen whether this practice will also restore fungal communities characteristic of these forests.

Ectomycorrhizal fungi amounted to nearly one-quarter of the reads in decomposing roots, a pattern of decomposition very different from that for surface litter. The asymptotic decomposition often seen during fine-root decay may be a result of competition between saprotrophs and ectomycorrhizae, as well as the limited decomposer capacity of many ectomycorrhizal fungi. The shifts in community composition through stand development and decay stage warrant further autecological or whole-genome sequencing work to elucidate whether between-species variation in genetic capacity influences below-ground carbon cycling.

MATERIALS AND METHODS

Site description.

Samples were collected at the Silviculture Treatments for Ecosystem Management in the Sayward (STEMS) research site near Campbell River, British Columbia (www.for.gov.bc.ca/hre/stems). Two sites were studied: STEMS 1 (50°04′32.7″N, 125°25′10.8″W) was harvested in 2001, and STEMS 3 (50°03′15.5″N, 125°35′18.7″W) was harvested in 2008. Both sites were replanted following harvest with Douglas fir (Pseudotsuga menziesii) and western red cedar (Thuja plicata) and are located in the Very Dry Coastal Western Hemlock biogeoclimatic subzone, with STEMS 1 in a slightly drier variant (CWHxm1) than STEMS 3 (CWHxm2) (61). Soils at both sites are humo-ferric podzols, are of sandy-loam texture, and have moder humus form (62). British Columbia's Biogeoclimatic Ecosystem Classification scheme categories sites into similar units based on climate, vegetation, and soil characteristics. Although we cannot rule out spatial heterogeneity as a driver of fungal community composition, our sites are within the same biogeoclimatic subzone, and every effort was made to ensure site comparability on the basis of climate, vegetation, and soil characteristics.

At each site, four treatments were established: aggregate retention (AR), dispersed retention (DR), clear-cut (CC), and an uncut control (UC). The AR treatment consists of 0.02- to 0.3-ha retention patches within a larger clear-cut matrix (STEMS 1, 25.5 ha; STEMS 3, 35.0 ha), whereas DR consists of uniformly distributed retention trees at an original target density of 45 stems per hectare over a treatment area of 18.2 (STEMS 1) or 37.8 (STEMS 3) ha (60). Clear-cut areas were 10.9 and 14.9 ha, and uncut areas were 12.0 and 17.0 ha (for STEMS 1 and 3, respectively). The understory consisted primarily of Gaultheria shallon, Vaccinium parvifolium, and Mahonia nervosa, while the herb layer was dominated by Pteridium aquilinum, Polystichum munitum, Rubus ursinus, and Linnaea borealis.

Experimental setup.

Fine roots (<2 mm) were sourced from nursery-supplied Douglas fir seedlings (PRT Campbell River), dried at 60°C for 48 h to constant mass, weighed (0.5 ± 0.05 g), and placed in sealed 15- by 15-cm nylon mesh (0.5-mm) bags. At each site, five sampling locations were selected for the CC and UC treatments, at the edge of four retention patches for the AR treatments (AR-E), or at half the distance from the dripline of four retention trees in the DR treatments (DR-05) (Fig. 1A). Four additional “outside-area” sampling locations in the retention treatments were selected at least 10 m away from retention trees/patches to test for any spatial influence of VRH (AR-OA and DR-OA) (discussed in the supplemental text); these samples were used as additional replicates in the stand age analysis (see Fig. 1 legend). At each sampling location, a trench (approximately 0.5 m long by 0.2 m wide by 0.2 m in diameter) was dug, and litterbags were buried horizontally at the transition between the LFH/Ae and B horizons (Fig. 1B). Where soil was disturbed due to logging, samples were buried at 10-cm depth. Soil and LFH material were replaced after the bags were buried. The total number of litterbags considered in this study was 104. Litterbags were buried in late June 2014, and the bags were recovered after 43 and 360 days. Root samples were immediately frozen in liquid nitrogen and kept on dry ice before storage at −80°C for later processing.

DNA extraction, internal transcribed spacer (ITS), amplification, and sequencing.

DNA from root material was extracted with the MoBio (Carlsbad, CA, USA) DNA elution accessory kit after RNA extraction using the RNA PowerSoil total RNA isolation kit. Quality and initial DNA quantification were determined using a NanoDrop (Wilmington, DE, USA) spectrophotometer. Extraction contaminants (suspected carbohydrate and phenol carryover) resulted in the loss of 19 samples, 16 from the 43-day data set and 3 from the 1-year data set, with n = 2 as the largest reduction in sample size. Limited material prevented sample reextraction.

For fungal molecular identification, the ITS2 region of the fungal ribosomal operon was amplified following a protocol similar to that described by Clemmensen et al. (63), with the forward primer gITS7 and the reverse primer ITS4, targeting a 220- to 500-bp amplicon (64). Both forward and reverse primers contained a 10-nucleotide tag unique to each sample. Dual barcoding was used to control for the tag-switching phenomenon (65). PCR amplification was conducted in a Bio-Rad MJ Mini thermal cycler (Bio-Rad, Mississauga, Canada) in 50-μl reaction mixtures using 25 μl of DreamTaq Green master mix (Thermo-Fisher Scientific, Waltham, MA), 2.5 μl gITS7 (0.5 μM), 1.5 μl ITS4 (0.3 μM), 2 μl template DNA (5 ng · μl⁻¹), and 19 μl of nuclease-free water. PCR was performed for each sample in triplicate under the following conditions: 5 min at 94°C; 25 cycles of 30 s at 94°C, 30 s at 56°C, and 30 s at 72°C; and 7 min at 72°C. Negative controls and PCR products were visualized on an electrophoresis gel to confirm the presence of product. Triplicate PCR products were purified using QIAquick PCR purification kits (Qiagen, Venlo, Netherlands). PCR product concentrations were calculated using the Quant-iT PicoGreen doubles-stranded DNA (dsDNA) assay (Life Technologies Corp., Carlsbad, CA, USA), and equimolar amounts were pooled and normalized from each technical replicate before submission to the sequencing facility. Amplicons were sequenced on a PacBio-RS II system at the University of Washington PacBio sequencing facility (Seattle, WA, USA) in two runs, using 4 or 6 SMRT cells for the 43-day and 1-year libraries, respectively. PacBio circular consensus sequencing technology achieves error rates similar to those with the Illumina MiSeq platform for small (<2,500-bp) amplicons (66). PacBio was chosen to prevent read length bias for the typically longer and variable length of ITS amplicons (64). Validation of PacBio technology for fungal ITS amplicon sequencing has recently been provided by Tedersoo et al. (66).

Sequence processing, and species/functional assignment.

SCATA (http://scata.mykopat.slu.se/) was used to quality-filter and cluster sequences into OTUs corresponding to species at 98.5% sequence similarity (single-linkage clustering). Only reads with an average quality score above 20 and with single-position quality scores above 10 were kept, all reads with missing primer sequences or tags were removed, and any read <200 bp in length was discarded. For sequence clustering, the minimum length of pairwise alignment was set to 85% of the longest sequence, and a scoring function with a mismatch penalty of 1, a gap open penalty of 0, and a gap extension penalty of 1 was used. Homopolymers were collapsed to 3 bp before clustering as recommended previously (66, 67). The UNITE database (http://unite.ut.ee) (68) was included during clustering to identify OTUs to species. Chimeras were identified and removed from the data set by employing the UCHIME algorithm against the most abundant genotype from each cluster from the SCATA output (69). As well, each OTU was further individually inspected for authenticity and reliability as outlined by Nilsson et al. (70). The BLASTn algorithm and the UNITE database were then used to identify all unknown clusters to species (at 98.5% similarity). Clusters identified to species or genus were assigned to the functional guild “saprotrophs,” “ectomycorrhizal,” “ericoid mycorrhizal,” “molds,” “yeasts,” “root endophytes,” or “plant pathogens.” Functional assignment was established using published literature and the FUNGuild annotated database (71) (http://funguild.org). Guild assignment was conservative; if no published literature could identify a species or genus to guild, or if the confidence ranking in FUNGuild was below “probable,” the functional assignment was left unknown.

Statistics.

All analysis was conducted in R (v3.3.2), and results were considered significant at a P value of <0.05. The effects of harvest treatment (H1), decay stage (H2), and stand age (H3) on multivariate species composition were investigated with distance-based redundancy analysis (72) based on Bray-Curtis similarities using the capscale function in vegan (v.2.4.1) (73). We chose not to present rarefied results, as this would have resulted in low sampling depth (459 reads) or sample removal. Data loss via rarefaction has been linked to increased uncertainty through randomization leading to inflation of type I and II error rates (74). That said, we found similar results in db-RDA and PERMANOVA analyses of unrarefied and rarefied data (not shown), but we present unrarefied results for the above-mentioned reasons. Raw abundances were fourth-root transformed to reduce the influence of a few highly abundant species (75). For H3, AR-E and DR-05 samples were excluded from the analysis to remove any influence of mixed-aged sampling locations, but outside-area samples were included as additional replicates for the 6- and 13-year-old stands. Multivariate hypotheses were tested using permutational multivariate analysis of variance (PERMANOVA) (function adonis in vegan, 999 permutations). As PERMANOVA is sensitive to unequal multivariate dispersion, multivariate group means were tested for homogeneity of variances via the betadispser function in vegan. When significant PERMANOVA main effects were observed, post hoc pairwise contrasts were conducted using the pairwise.perm.manova function in the RVAideMemoire package (76).

Multivariate generalized linear models were used to evaluate individual species or guild response to factors using the mvabund package (v3.11.9) (77). These models complement PERMANOVA by providing an overall multivariate response to factors in addition to univariate response. All results were conducted assuming a negative binomial distribution and resampled 999 times using the PIT-trap method, and the likelihood ratio was used as the test statistic. Hypothesis testing was conducted using the anova.manyglm function, and treatment contrasts were evaluated using the summary.manyglm function. P values were adjusted for multiple testing using a step-down resampling procedure. For the species data, only the globally most abundant taxa were analyzed, representing at least 90% of the summed mean relative abundance and a minimum within-sample relative abundance of 80%. Z-scores of significant taxa abundance were visualized using ggplot2.

Accession number(s).

All sequences were deposited with the NCBI under accession number SRP118187.