Author for correspondence: Roger T. Koide Tel: +1 814 863 0710 Fax: +1 814 863 6139 Email: firstname.lastname@example.org
• Several mechanisms may contribute to the high species richness often reported in ectomycorrhizal (ECM) fungal communities, including spatial and temporal partitioning. Here, we focus on temporal partitioning.
• Using molecular methods, we determined the frequencies of occurrence of ECM fungal species detected as hyphae and ECM roots in the forest floor of a Pinus resinosa plantation during a 13-month period. We then used a novel statistical procedure to place the most frequently occurring ECM fungal species into groups distinguished by their patterns of relative frequency over time.
• Three groups with contrasting temporal patterns were distinguishable for fungal species detected as hyphae. Two groups were distinguishable for species detected as ECM roots.
• Our results support the hypothesis that temporal partitioning occurs among the species of ECM fungi in this community, but we did not address its causes, which may have involved interactions among species’ physiological tolerances, temporal environmental variability, temporal patterns of root production, and variation in fungal genet lifespan. These interactions should be the subjects of future research.
By temporal partitioning we are not referring to successional changes in species composition occurring over periods exceeding a year (Mason et al., 1983; Visser, 1995; Nara et al., 2003), which is evidence for the absence of stable species coexistence. Instead, temporal partitioning is a mechanism that promotes stable species coexistence by reducing the likelihood of competitive exclusion, such as when species are active at different times of year. Such temporal partitioning by species or species groups occurs within soil bacterial communities (Bossio et al., 1998; Smit et al., 2001), but there is little direct evidence for temporal partitioning within ECM fungal communities. There are many observations that species of higher fungi fruit at different times of the year (Giachini et al., 2004). However, we are unaware of any study documenting contrasting intra-annual patterns of activity of ECM fungi.
Here we report on a test of the hypothesis that temporal partitioning exists within an ECM fungal community. We tested this hypothesis by analyzing frequencies of occurrence of ECM fungi detected as ECM roots and as hyphae in the forest floor over the course of a 13-month period using a novel statistical procedure to group species according to their temporal patterns.
Materials and Methods
In this study, frequencies of occurrence of ECM fungal species were determined for forest floor hyphae and for ECM roots on seven sampling dates over the course of 13 months. The data used in the analyses presented here were also used in previous studies (Koide et al., 2005a,b). In those studies, however, the data from all seven sampling dates were pooled and then analyzed for the purpose of either documenting interactions among species (Koide et al., 2005a) or determining whether there was any relationship between a species’ frequency as hyphae and as ECM roots (Koide et al., 2005b). In the current study, we have kept separate the data from each sampling date in order to analyze the temporal patterns of activity of the component species of ECM fungi. Because much of the methodology for the current study is the same as in previous studies (Koide et al., 2005a,b), we include here only a general description of the study site, the additional methods pertinent to this study, and a few points to further clarify previously reported methods.
The study was performed in a closed canopy stand of an c. 65-yr-old red pine (Pinus resinosa Ait.) plantation located in State College, Centre County, PA, USA. An analysis of four, 7-m-radius plots within the plantation revealed that the mean stand density was 1590 trees ha−1 (SE 110), the mean basal area was 45.3 m2 ha−1 (SE 1.2), and the mean tree diameter was 22.0 cm (SE 1.0). The site has a well-developed O-horizon comprising litter, a fermentation (F) layer, and a thin humified (H) layer, which overlies the mineral soil (Morrison sandy loam), consisting of a thin eluviated A-horizon over a well-defined, sandy B-horizon. Sampling was restricted to a contiguous area without understory vegetation, approx. 2.7 ha in extent. Air temperatures and rainfall data were obtained from AccuWeather data for State College (http://wwwa.accuweather.com/index).
Sampling of fungi
Hyphae of ectomycorrhizal fungi Two-centimetre (inside diameter) cores of F-layer were collected from among 80, randomly located 0.25 m2 plots on 18 July, 16 September and 16 October 2002, and 16 January, 5 April, 26 June and 22 August 2003. On each date, 60 of the 80 plots were randomly selected for coring for a total of 420 F-layer samples. In the field, we took unhomogenized 1.0 ml samples of the F-layer from each core, placed them in microfuge tubes, and kept them frozen (−20°C) until extracted for DNA analysis.
Ectomycorrhizal roots The small cores taken for hyphal sampling did not reliably contain ECM roots. Therefore, these were collected from cubes (5 × 5 cm surface) of F-layer removed from the forest floor with a small knife within a few cm of the cores, within the same plots and on the same dates. On the first four dates, 40 out of the 80 plots were randomly sampled. On the subsequent three dates, 60 of the 80 plots were randomly sampled. Thus, a total of 340 cubical F-layer samples were removed for analysis of ECM roots. Three ECM roots (for the first six dates) or four ECM roots (for the last date) were randomly chosen from among all ECM roots of each sample and individually dried over silica gel in microfuge tubes. We previously showed that there was no significant difference between sampling three or four ECM roots per cube (Koide et al., 2005b). Ectomycorrhizal roots were then stored individually in microfuge tubes at −20°C until they were extracted. A total of 1080 ECM roots were thus individually processed.
Identification of fungal species
Fungal species were identified as described in previous studies (Koide et al., 2005a,b) by terminal restriction fragment length polymorphism (T-RFLP) analysis following PCR amplification of genomic DNA (ITS-1F, Gardes & Bruns, 1993, labeled with VIC, and ITS4, Gardes & Bruns, 1993, labeled with NED), restriction digestion with Hinf1 and HaeIII, and terminal restriction fragment analysis using a capillary electrophoresis DNA sequencer.
Terminal restriction fragment lengths were compared with those in our database comprising ECM fungi collected from our research site over several years as sporocarps or ECM roots. For ECM roots, we considered a species present if its four restriction fragments (VIC Hinf1, NED Hinf1, VIC HaeIII, and NED HaeIII) were observed in the sample T-RFLP fingerprint. For hyphal samples, we considered a species present if at least three of its four restriction fragments were observed. The reason for this is that, in the hyphal samples, we frequently found that there were many short peaks in addition to a few taller peaks in the T-RFLP profile (e.g. see Fig. 1 in Dickie et al., 2002). These short peaks probably represent terminal restriction fragments of various fungi, both mycorrhizal and saprotrophic, which were poorly amplified, probably because of low amounts of extractable genomic DNA. On occasion there were so many of these small peaks that it was possible to pick out nearly any T-RFLP fingerprint that one could want. In order to eliminate the possibility of wrongfully observing fingerprints of fungi that were not really present, we disregarded peaks of <50 fluorescence units. In most cases the four expected restriction fragments for a given ECM fungal species remained, but on occasion, one restriction fragment would be eliminated by our ‘minimum’ criterion. We judged this approach to be more conservative than one in which we included all peaks irrespective of their size.
In the vast majority of cases, each restriction fragment in a fingerprint was matched with the database value to within 1.0 bp error. We usually found that the sporocarps used to construct the database produced the expected four terminal restriction fragments, each of a unique length. When no restriction site existed for one of the restriction enzymes, the PCR product length was recorded as the restriction fragment length for that enzyme. Our database includes ECM fungal species and species identified repeatedly from healthy ECM roots (designated Jori #) and thus assumed for the time being to be mycorrhizal. A few sporocarps produced three restriction fragments, each of unique length, and a fourth restriction fragment of two lengths, 2 bp apart, seen as a peak doublet in its T-RFLP profile. An unknown hyphal or ECM root sample might produce only one of the peaks of the doublet representing that fourth restriction fragment. In order to accommodate this case, we considered the unknown to match with a species in the database if the three single peaks matched within 1.0 bp and if the fourth fragment of the unknown was within 1.0 bp of either of the doublet peaks in the database.
For ECM fungal species detected as hyphae, the DNA analysis was only capable of determining species presence or absence irrespective of the amount of hyphae of that species in the F-layer sample. Thus, for the species detected as ECM roots, the data were treated in the same fashion; an ECM fungus species was considered to be either present in or absent from each sampling unit (the F-layer cube), irrespective of the number of roots in the sample colonized by that species (Koide et al., 2005b). The need to consider the ECM root data in this fashion was discussed previously when we determined that the identity of the fungus on one root in a sample was not independent of the others in the same sample (Koide et al., 2005b). Therefore, for ECM fungal species detected either as hyphae or as ECM roots, this method gives equal weight to all species within a sample, irrespective of their abundance within the sample.
Because ECM roots were sampled either 40 or 60 times per sampling date, depending on the sampling date (as already described), species occurrences were calculated for both ECM roots and hyphae on a per-sample basis to make comparison among dates possible. For each species, these values were then scaled by dividing the value at each sampling date by the sum of the values for all seven sampling dates. Thus, for a given species, the sum of these relative frequencies for the seven sampling dates was always 1. The purpose of this scaling was to characterize the temporal pattern of occurrence for each species in a manner that was independent of its absolute frequency of occurrence and independent of all other species in the community.
Evidence for temporal partitioning among the species of ECM fungi was obtained using PROC TRAJ, a semiparametric procedure that fits models to longitudinal data in order to produce data-sequence groupings (see the PROC TRAJ website: http://www.andrew.cmu.edu/user/bjones/; see also Nagin, 1999, 2005; Nagin & Tremblay, 2001). In our case, we used PROC TRAJ to group species according to their similarity in relative frequency through time. Because PROC TRAJ was applied to the relative frequencies as calculated earlier, it separated species into groups according to temporal pattern only (relative frequency through time), and not on the basis of absolute frequency of occurrences, which varied greatly among species. We analyzed only the 14 most frequent species detected as hyphae (out of a total of 28; Koide et al., 2005b) and the seven most frequent species detected as ECM roots (out of a total of 27; Koide et al., 2005b). The minimum criterion for inclusion in the analysis was 15 occurrences throughout the sampling period. For less common species, relative frequencies would be particularly sensitive to chance fluctuations in occurrence as a result of sampling and detection error.
We employed censored normal models in PROC TRAJ, initially specifying third-order polynomial models. If PROC TRAJ indicated that the cubic or quadratic parameters were not significant, reduced models (linear or quadratic rather than cubic) were specified. We present results from the reduced models only. One of the peculiarities of PROC TRAJ is that all relative frequencies must be expressed to the same degree of precision. In other words, frequencies of 25 of 100 and 33 of 67 must not be expressed as 0.25 and 0.4925, but rather as 0.2500 and 0.4925, or as 0.25 and 0.49. For the analyses herein, we arbitrarily expressed all relative frequencies to four decimal places. The results would have been the same had we expressed them to other degrees of precision so long as this was consistent across all relative frequencies. In the model, the timescale (originally 199–599 d since 1 January 2002) was centered about 0 and scaled (between −0.36 to +0.36) to be comparable in magnitude to that of the dependent variable (frequency), which ranged approximately from 0.0000 to 0.3500, as recommended (D. S. Nagin, pers. comm.).
Correlation analyses were performed using the Pearson product-moment correlation procedure of Statistica version 5.1 (Statsoft, 1997) to determine whether relative frequency of each species observed on ECM roots was correlated with the relative frequency of the species as hyphae.
Rainfall for the 30 d before each sampling was lowest for the first two samplings and highest for the third, sixth and seventh samplings (Fig. 1a). Air temperatures were typical for the region, with the highest 30 d means of daily means near 23°C in midsummer and the lowest 30 d mean of daily means below freezing in midwinter (Fig. 1b).
There were a total of 39 fungal species determined to be ECM in our analyses. Twenty-eight species in total were detected as hyphae (Koide et al., 2005b), and 27 as ECM roots (Koide et al., 2005b). There were 10, 17, 16, 17, 16, 12 and 21 species of ECM fungi detected as hyphae in July, September and October 2002, and January, April, June and August 2003, respectively. For the same dates, there were 10, 10, nine, nine, 11, 14 and 17 species of ECM fungi detected on ECM roots.
As can be readily appreciated from the relative frequencies for species detected as hyphae (Fig. 2), it would be difficult to objectively group species according to temporal pattern without the aid of a computer. Thus, for the species detected as hyphae, we specified that PROC TRAJ produce three, four and five groups. Based on the suggested diagnostic procedure (Nagin, 2005), all three models were adequate because each produced average posterior probabilities of group membership of 0.90 or higher. However, both the Akaike information criterion (AIC) and Bayesian information criterion (BIC) indicated that the three-group model was most parsimonious and thus we report the results only from the three-group modeling effort. The same reasoning was used in our selection of the two-group model for ECM root data based on comparisons with three- and four-group models.
For species detected as hyphae, Group 1 consisted of Lactarius oculatus, Suillus intermedius, Tylopilus felleus and Jori 10, which generally exhibited the highest frequencies in the autumn and the lowest frequencies in the spring (Fig. 3a). Group 2 consisted of Amanita cf. vaginata, Jori 1, Jori 4, Jori 5 and Ramaria concolor, which generally exhibited the highest frequencies in the winter or spring, and their lowest frequencies in mid- to late summer (Fig. 3b). Group 3 consisted of Amanita brunnescens, Cenococcum geophilum, Clavulina cinerea, Russula white 1 and Scleroderma citrinum, which exhibited little variation in relative frequency throughout the sampling period (Fig. 3c). For clarity, the models for the three groups produced by PROC TRAJ are shown in Fig. 4.
For the species detected as ECM roots, Group 1 comprised C. geophilum, C. cinerea, L. oculatus, Russula white 1 and T. felleus (Fig. 5a). This group exhibited little variation in relative frequency throughout the sampling period. Group 2 comprised A. brunnescens and Jori 4, which exhibited the highest frequencies in spring or summer, and the lowest frequencies in autumn (Fig. 5b). Because there were only two groups, the group models are not shown separately.
There were no significant correlations for any species between relative frequency as hyphae and relative frequency as ECM roots. However, most of the species that exhibited little variability in relative frequency as hyphae were among those that exhibited little variability as ECM roots, including C. geophilum, C. cinerea, Russula white 1 (compare Figs 3b and 5a). Likewise, the temporal pattern of Jori 4 detected as hyphae was similar when detected as ECM roots (compare Figs 3b and 5b). In contrast, for A. brunnescens, L. oculatus and T. felleus, the relative frequency patterns for hyphae and ECM roots were different (compare Figs 3a,c and 5a,b). Thus, the temporal patterns of relative frequency for the various species were not necessarily the same when detected as hyphae and when detected as ECM roots.
We present here some of the first evidence for temporal partitioning among species of an ECM fungal community. Groups of ECM fungal species, detected as either hyphae or ECM roots in the F-layer, exhibited distinctive temporal patterns in their relative frequencies of occurrence. For those detected as hyphae, one group of species showed little variation in relative frequency throughout the year. Another group exhibited maximal relative frequency in autumn and minimal relative frequency in the spring. A third exhibited nearly the opposite pattern, with maximal relative frequency in winter or spring and minimal relative frequency in mid- to late summer.
The hyphae of all species of ECM fungi obtain resources such as water, nitrogen (N), phosphorus (P), and other nutrient elements from the forest floor. Temporal partitioning by hyphae in the forest floor may thus serve to reduce the amount of competition for these resources among species in different groups, reduce the likelihood of competitive exclusion, and thus help to maintain species richness. The maintenance of species richness by this and other mechanisms may be important in maintaining a diversity of functions in the community, because, for example, substantial variability occurs among species in their ability to utilize various N sources (Abuzinadah & Read, 1986; Arnebrant, 1994; Keller, 1996; Dickie et al., 1998), in their response to water stress (Coleman et al., 1989), and in their ability to transport materials through rhizomorphs (Read, 1991; Rousseau et al., 1994; Agerer, 2001).
Species detected as hyphae possessing the same temporal pattern might be expected to compete more amongst themselves than with species with contrasting temporal patterns. However, such species may stably coexist if they are distinguished in other ways, such as in space. Indeed, while C. geophilum, C. cinerea, A. brunnescens, Russula white 1 and S. citrinum all occurred in the same temporal pattern group, we previously showed that significant negative interactions occurred on a fine spatial scale (the scale of the sample) between C. geophilum and C. cinerea, S. citrinum and A. brunnescens, and C. cinerea and Russula white 1 (Koide et al., 2005a). By contrast, Jori 1, Jori 4 and Jori 5 were all members of the same temporal pattern group, yet Jori 1 was positively correlated with both Jori 4 and Jori 5 in space (Koide et al., 2005a). This suggests that these species do not occur in a high enough abundance to compete strongly, are in the process of competitively excluding each other, or that other factors such as selective fungivory or interannual variation in climate serve to prevent local extinction. Further investigations into the relationships among spatial partitioning, temporal partitioning and other factors may prove to be helpful in testing hypotheses about species coexistence and competition.
There was no clear evidence for temporal partitioning of the species when they were detected as ECM roots. One group of species (comprising five of the seven species that were analyzed) possessed a fairly constant relative frequency throughout the sampling period. Another group (two of the seven species) exhibited maximal relative frequencies in spring to summer and minimal relative frequencies in autumn. We detected no species exhibiting a complementary pattern to that. Nevertheless, the existence of two distinct patterns of relative frequency certainly suggests that temporal partitioning of roots by ECM fungal species could occur. If it did, the effect could be a reduction in the likelihood of exclusion as a consequence of competition for host photosynthate. The timing of new root production could influence the temporal patterns of the species detected as ECM roots because the various species did exhibit different patterns of hyphal activity throughout the year. Nevertheless, it is difficult to conclude much about the relationship between hyphal activity and temporal patterns of root colonization in this study because we do not know whether different fungal species influence root longevity differently. One would have to know the age of roots at the time of their collection.
Out of a total of 39 species of ECM fungi that we detected, we only analyzed seven species detected as ECM roots and 14 species detected as hyphae, because we judged that species of lower frequency could not reliably be grouped according to temporal pattern. Of course, this is not a limitation of PROC TRAJ per se, but rather a consequence of the structure of most communities of mycorrhizal fungi, which comprise a few common species and many rare ones (Horton & Bruns, 2001). More frequent sampling during the 13-month period, more intensive sampling, or a longer sampling period would have been useful in further delimiting species groups, and future investigators should take that into consideration.
The use of PROC TRAJ to form species groups could in some cases mask some variability in temporal pattern because some degree of heterogeneity will always exist within multispecies groups. Nevertheless, the placement of multiple species into groups may allow one to perceive large-scale patterns that are not observable if each species were considered separately. In this respect, the assembly of groups by PROC TRAJ is similar to the construction of functional groups in studies of ecosystem function (Bengtsson, 1998). The major distinction, however, is that PROC TRAJ does not assume a priori assignment of species into groups. Instead, it groups species only after an examination of their frequencies over time. Moreover, it is possible for PROC TRAJ to place a species into a group of its own if its behavior differs significantly from that of others. Therefore, the use of PROC TRAJ does not necessarily result in any undue loss of information through the aggregation of species.
It was not our purpose to investigate the causes of temporal partitioning by the various species of ECM fungi. However, we can at least suggest some of the possible causes of the observed temporal patterns. First, it is clear that different species have different preferences for temperature and moisture (references cited earlier). Cenococcum, for example, has been shown to be particularly drought-resistant (Mexal & Reid, 1973; Pigott, 1982; Coleman et al., 1989; Jany et al., 2003). One therefore expects to encounter it more frequently relative to other, less drought-resistant species following periods of drought and, in communities in which drought occurs periodically, one expects Cenococcum to maintain a relatively high frequency throughout the year, as it did. Other species, which are not as drought-tolerant, might be expected to be temporally less stable than Cenococcum in relative frequency. Second, because roots are produced under a variety of environmental conditions (Fahey & Hughes, 1994; Rygiewicz et al., 1997; Thomas et al., 1999; Burton et al., 2000; Johnson et al., 2000; King et al., 2002; Tierney et al., 2003), different cohorts of roots may be colonized by different groups of fungi, depending on their tolerance to the conditions under which the roots became available. Finally, even if different species of ECM fungi colonize roots at the same time of year and have the same general temporal pattern of activity, variation among species in the lifespans of their genets, which can be substantial (Dahlberg & Stenlid, 1995; Zhou et al., 2000; Fiore-Donno & Martin, 2001; Zhou & Hogetsu, 2002; Guidot et al., 2003), could result in distinct temporal patterning.
We gratefully acknowledge funding from the A. W. Mellon Foundation, and from the National Research Initiative of the USDA Cooperative State Research, Education, and Extension Service (Competitive Grant no., 2002-35107-12243). We thank K. B. Boomer and T. Wrobel of the Statistical Consulting Center, The Pennsylvania State University, for their assistance with PROC TRAJ, and I. Alexander and two anonymous referees for their considerable assistance in revising this manuscript.