Interactions among fungal community structure, litter decomposition and depth of water table in a cutover peatland


  • Editor: Karl Ritz

Correspondence: Clare J. Trinder, Institute of Biological and Environmental Sciences, University of Aberdeen, Cruickshank Building, St Machar Drive, Aberdeen AB24 3UU, UK. Tel.: +44 1224 272692; fax: +44 1224 272703; e-mail:


Peatlands are important reservoirs of carbon (C) but our understanding of C cycling on cutover peatlands is limited. We investigated the decomposition over 18 months of five types of plant litter (Calluna vulgaris, Eriophorum angustifolium, Eriophorum vaginatum, Picea sitchensis and Sphagnum auriculatum) at a cutover peatland in Scotland, at three water tables. We measured changes in C, nitrogen (N) and phosphorus (P) in the litter and used denaturing gradient gel electrophoresis to investigate changes in fungal community composition. The C content of S. auriculatum litter did not change throughout the incubation period whereas vascular plant litters lost 30–40% of their initial C. There were no differences in C losses between low and medium water tables, but losses were always significantly less at the high water table. Most litters accumulated N and E. angustifolium accumulated significant quantities of P. C, N and P were significant explanatory variables in determining changes in fungal community composition but explained <25% of the variation. Litter type was always a stronger factor than water table in determining either fungal community composition or turnover of C, N and P in litter. The results have implications for the ways restoration programmes and global climate change may impact upon nutrient cycling in cutover peatlands.


Peatlands are estimated to cover c. 400 million ha of land worldwide (Moore, 2002) and may contain as much as 455 Pg of carbon (C) (Gorham, 1991), thus representing an important reservoir of C. Demand for peat for horticultural use and large-scale power generation has increased and new mechanized extraction methods have facilitated a dramatic increase in exploitation of peatlands. In the European Union alone, it has been estimated that 38% of the total area of peatland has been severely affected by commercial harvesting (Raeymaekers, 2000). At the end of harvesting, many sites are abandoned and this can result in significant alterations to the C balance of peatlands. In some cases, they can change from acting as a net sink to a net source of C to the atmosphere (Waddington et al., 2002), while afforestation of many areas (Holden et al., 2007) has uncertain consequences for the C balance.

Even in the absence of restoration, cutover peatlands will generally be colonized by acidophilic plant species. This colonization process is critical to peatland biogeochemical cycling because plants provide the only input of new C to these systems. The rate at which litter decomposes will have a fundamental effect on whether peat continues to develop or the site becomes a net source of CO2. Litter chemistry varies among plant species and is known to affect rates of decomposition; in impoverished environments, plants tend to produce litter with low concentrations of key elements such as phosphorus (P) and nitrogen (N) with consequently high C : N quotients. The C : N quotient is considered an important indicator of the decomposability of plant litter. The theoretical optimum C : N quotient for microbial growth is 30 (Deacon, 1997); if the C : N quotient is greater than this, microorganisms may immobilize N from other sources, which reduces the imbalance of C and N and enables decomposition to continue (Heal et al., 1997). Sphagnum mosses are key peat-forming species; like other nonvascular plants, they contain phenolic compounds, waxes and polymerized lipids rather than lignin (Aerts et al., 1999); polyphenols and lignin are resistant to microbial decay, especially in anaerobic conditions (Brown, 1997). Unrestored cutover peatlands may have very little Sphagnum cover; hence, rates of litter decomposition from other plants now found there will be important in determining whether peat starts to develop again.

A second crucial regulator of the rate of litter decomposition is the diversity and structure of the microbial community. Fungi are thought to be more important than bacteria in terms of litter decomposition (Thormann, 2006). In common with plants colonizing newly exposed substrates, decomposer fungi in peatlands are thought to show a succession of species (or types) colonizing plant litter from senescence to its complete decomposition (Thormann, 2006). Cutover peatlands tend to contain lower microbial biomass than intact mires because the peat exposed at the surface is often of ancient origin, subject to extreme fluctuations in moisture content and temperature and the C is highly recalcitrant (Croft et al., 2001). Most studies to date have used culturing techniques to gain insight into microbial communities in peat; for example, comparing cutover peatlands with undamaged sites (Dooley & Dickinson, 1970), studying succession on litter of different peatland plants (Thormann et al., 2003), differences between decomposition of leaf and root material (Thormann et al., 2001) and correlations between microfungi and plant tissue variables such as P and N (Thormann et al., 2004). However, it is now widely acknowledged that only a small proportion of soil microorganisms can be cultured (Bridge & Spooner, 2001); hence, these studies provide partial information about the roles of microbial biodiversity for biogeochemical cycling in peatlands. In one of the few molecular studies undertaken to date, Artz et al. (2007) compared measures of fungal diversity using denaturing gradient gel electrophoresis (DGGE) and sequencing, with those based on culturable microorganisms. This analysis provided clear evidence that culture techniques generally record primarily fast-growing fungal species that produce large numbers of spores.

A third factor of considerable importance in regulating litter decomposition in peatlands is the underlying hydrological conditions. Some climate models predict that northern peatlands are likely to become drier and warmer in summer conditions (Mitchell, 1989), which has the potential to have a significant effect on microbially driven processes in these ecosystems (Martikainen et al., 1993). In addition, peatland restoration programmes often involve raising water levels to restore appropriate conditions for the reestablishment of Sphagnum spp., and thus recommencing peat formation (Holden, 2005). Higher water tables generally reduce rates of decomposition of plant material (Clymo, 1983) and few fungi can tolerate anaerobic conditions for extended periods (Deacon, 1997). Proximity of plant litter to the water table is, therefore, an important determinant in rates of decomposition. However, although rates of decay are faster at drained sites (Bridgham & Richardson, 2003), very dry conditions can also reduce fungal growth (Laiho, 2006; Jaatinen et al., 2007). It is unlikely that appropriate hydrological conditions for reestablishment of Sphagnum can be restored on all cutover peatlands – such methods are often costly (Holden, 2005) and may have adverse effects on surrounding land, even where there is an interest in undertaking such work. Regardless of whether peatland hydrology is affected by climate change or restoration, there remains a crucial gap in our understanding of the combined effects of changes in water table depth and the composition of litter inputs from different colonizer plant species on fungal diversity and nutrient turnover.

Here, we used molecular fingerprinting to investigate changes in the community structure of fungi colonizing plant litter on a cutover peatland. The effects of different moisture regimes on the litter decomposition rates of typical colonizing plant species and structure of their microfungal populations have not previously been studied in the context of cutover peatlands. We used five species commonly found on abandoned cutover peatlands: Calluna vulgaris (L.), Eriophorum angustifolium (L.), Eriophorum vaginatum (Roth.), Picea sitchensis (Bong.) and Sphagnum auriculatum (Schimp.), and investigated changes at three different water tables in an 18-month experiment. In addition, we correlated changes in fungal community composition with changes in litter chemistry. We hypothesized that: (1) distinct communities of fungi would be associated with different litter types and with different water tables and these would show successional shifts over subsequent harvests; (2) changes in litter chemistry would be correlated with fungal community structure; (3) increasing the height of water table would reduce losses of C and P; and (4) N would accumulate in litter during the early stages of decomposition.

Materials and methods

Study site

The study was carried out at Middlemuir Moss, a former raised mire site in North East Scotland: National Grid Reference NJ 906563, altitude 110 m a.s.l. The site has a long history of manual and mechanized peat cutting; it is estimated that a depth of up to 4 m of peat has been removed during mechanized harvesting operations that started in 1961. Peat was harvested in baulks and trenches c. 100 m wide and the remaining peat depth varies from 2 to 3 m. The peat is highly acidic (pH≤3) and humified. No restoration has been carried out and the site is unmanaged. Much of the site is dominated by C. vulgaris, E. vaginatum and E. angustifolium with scattered patches of Sphagnum spp. and extensive areas of bare peat.

The experiment used a water table gradient created during earlier harvesting operations. It measured c. 100 m2 and was divided into three blocks. The water table gradient ran perpendicular to the block structure so that each block contained a high, medium and low water table. Over a year, the high water table averaged 10 ± 2.5 cm above the ground surface, medium water table averaged 9 ± 2 cm below the ground surface and the low water table averaged 18 ± 3 cm below the ground surface. All the water table depths were significantly different from each other (P<0.05). The area was cleared of vegetation 1 year before the start of the experiment and was maintained free of vegetation throughout the experiment.

In November 2004, plant material from E. vaginatum, E. angustifolium, C. vulgaris and S. auriculatum was collected from the site. Only leaves and stems that had already lost their chlorophyll and senesced were collected. This ensured that litter would be chemically and physically similar to that falling from the plants while reducing effects due to decomposition once litter has fallen. For S. auriculatum, growing plants were collected and the top 1 cm of the plant (including the capitulum) was discarded. Needles of P. sitchensis, still attached to branches, were collected from another site (National Grid Reference: No. 871885). All plant litter was air dried to constant weight and a subsample of each litter type was frozen for later analysis. A total of 270 litter bags were prepared using 500 μm mesh (Plastok Associates Ltd, Birkenhead, UK); each litter bag held 4 g of litter of one litter type only and was glued closed. Owing to different densities of the plant litter, litter bags measured 12 × 6 cm (E. angustifolium, P. sitchensis and C. vulgaris), 15 × 6 cm (E. vaginatum) and 24 × 6 cm (S. auriculatum). Fifty-four sets of litter bags were prepared, each set comprising one litter bag of each litter type, arranged in random order.

In December 2004, 18 sets of litter bags were buried at each water table (low, medium and high), evenly distributed across the three blocks. Litter bags were buried vertically to keep disturbance of the surface layers of the peat to a minimum, in 6-cm-deep slots, with their top edge flush with the ground surface. This was designed to mimic the dynamics of litter falling into cracks in the peat. Two sets of litter bags were selected at random from each water table in each block for harvest after 28, 56 and 80 weeks (referred to as Harvests 1, 2 and 3, respectively; starting conditions referred to as time 0), giving six replicates of each litter type at each water table at each harvest.

At each harvest, litter bags were removed and stored on ice during transport back to the laboratory, where material in the litter bags was removed and weighed. A subsample (0.7–0.8 g) was removed and frozen for DNA extraction. Remaining material was oven dried at 105 °C for up to 48 h. Material from Harvest 3 was subdivided and half of it was washed using distilled water before oven drying to determine how much peat had entered the litter bags and used to correct figures from the earlier harvests, assuming a linear increase in peat entering the litter bags over time. Plant litter was milled and C and N content was determined using an elemental microanalyser (Na 1500 NCS, Fisons Instruments, UK). P was determined using sulphuric acid/hydrogen peroxide digestion and measured using colorimetric determination at 882 nm after reaction with molybdenum blue (John, 1970). Certified reference material (Poplar leaves, NCS DC73350, China National Analysis Centre for Iron and Steel, 2004) was used as a standard in each digest and used to correct for losses during digesting. Time 0 samples were analysed simultaneously with Harvest 1 samples.

DNA extraction, PCR and DGGE

DNA extraction, PCR and thermal cycling conditions were carried out as described in Artz et al. (2007). Briefly, nucleic acids were extracted as described by Griffiths et al. (2000) from 0.2 g of litter and fungal DNA amplified using the primer pair ITS1F (Gardes & Bruns, 1993) and ITS4 (White et al., 1990). First-round PCR products were cleaned with Wizard SV PCR cleanup kit (Promega, UK). Purified PCR products were used as templates in a nested PCR using fungal-specific ITS1-GC and ITS2 (White et al., 1990) primer pair. To account for differences in banding patterns being affected by the water content of samples taken from different water tables, control samples were rewetted using distilled water to the same proportion as the samples removed from the field at different water tables at Harvest 1.

DGGE was carried out using the DCode Universal Mutation Detection System (Bio-Rad, UK) on 8% polyacrylamide gels bound to GelBond PAG film (Cambrex, UK) with a linear vertical gradient of 30% [2.0 M urea–12% (v/v) formamide] to 60% (4.2 M urea–24% formamide). Gels were cast using a gradient former (Fisher Scientific, UK) at flow rate of 4 mL min−1. Approximately 150 ng of nested PCR products were loaded onto the gels. Gels were run for 16 h in 1 × Tris-acetate–EDTA (TAE) buffer (40 mM Tris-acetate, 1 mM EDTA) at 75 V at 60 °C. Gels were stained with silver nitrate solution (McCaig et al., 2001). DGGE gels were standardized using a standard mixture of DNA fragments (Hyperladder IV, Bioline, UK). Gels were digitized by image scanning using an Epson GT-9600 scanner as uncompressed 16-bit greyscale TIFF files. Samples from each harvest were analysed together but within each harvest, samples were assigned randomly to gels; a total of 23 gels were run.

Fungal DNA extracted from harvested litter samples was compared with DNA extracted from the unburied litter samples used as starting material (time 0). DGGE gels were analysed using gelcomparii version 4 program (Applied Maths, Kortrijk, Belgium). Hyperladders were used to standardize among gels and band searching carried out as described in Artz et al. (2007), using settings that gave 99% similarities for standards. Data for presence or absence of bands in band classes, assuming a band represents a single fungal species, were exported as a binary matrix for statistical analyses.

To determine the variability of the PCR process and gel staining, following analysis of samples from the first harvest, one sample from each litter type showing a range of high and low band richness was selected. Five replicate samples were taken from the initial DNA extract and taken through the PCR process. They were randomly assigned and run on two DGGE gels as described above.

Calculations and statistical analyses

In all cases, data were analysed for each harvest separately to overcome interaction effects with time that can lead to difficulties interpreting significance of main effects (Wieder & Lang, 1982). For statistical tests, changes in C, N and P in litter were expressed as a percentage of the original amount present at the start of the experiment. Data were analysed in genstat (version 7) using anova following log transformations as appropriate to ensure data were normally distributed and homogeneous and taking into account the block structure of the experiment. Where results were significant, post hoc Tukey's tests were carried out. Variance partitioning was calculated from the sum of squares (SS) from the anovas by expressing SS for each variable as a percentage of total SS. Fisher's least significance difference (LSD) post hoc tests were calculated where there were significant interaction effects.

The effects of water table and litter type on the DGGE fingerprints were visualized following a principal coordinates analysis (PCoA) on the binary data matrix (Artz et al., 2007). Repeat samples and the standards run on the respective DGGE gels were analysed in a separate PCoA to determine how much variability was due to PCR effects and differences among gels. Tests for significance and variance partitioning of the effects of harvest, litter type and water table was carried out on the binary matrix using distance-based multivariate analysis for a linear model (distlm v.5: Anderson, 2001; McArdle & Anderson, 2001), with 9999 permutations using unrestricted permutation of the raw data and taking into account the blocking structure of the experiment. Post hoc tests are not available in this program.

Canonical correspondence analysis (CCA) was used to show similarity or differences between fungal community structure associated with the litter types and water tables, with the environmental variables superimposed over the fungal community composition data as vectors. The significance of the vectors is shown by their length and direction from the origin. The analysis was undertaken using canoco for Windows (version 4.5, Biometrics, the Netherlands) with Monte Carlo mean squares, 999 permutations and manual settings. Validity of using this model was tested using detrended correspondence analysis on each of the binary fungal community composition datasets for Harvests 1, 2 and 3. The gradient lengths of the longest gradients were 2.5, 4.6 and 3.0, respectively. Gradient lengths of >3.0 generally indicate that a unimodal method (such as CCA) is appropriate (Lepš & Šmilauer, 2003) and for consistency, we used this for Harvest 1 as well. Changes in C, N and P and fungal community structure from PCR fragments were included in the analyses as environmental variables and tested individually for percentage variance and statistical significance and again as part of a model using forward selection.


Litter quality

Table 1 shows the initial amounts of C, N and P in the different litter types and in the peat. Table 2 shows the percentage mass remaining of the different litter types at different water tables. Spearman's rank order correlation showed that percentage mass and C remaining were highly correlated (rs=0.964, P<0.001).

Table 1.   Concentrations (mg g dwt−1) of C, N and P in peat and in bulk litter before burial
 C (mg g dwt−1)N (mg g dwt−1)P (mg g dwt−1)
  1. Means ± SE; for peat samples, n=3; for litter types, n=6. For litter samples, SEs represent homogeneity of litter.

Peat545 ± 1618.8 ± 1.61.43 ± 0.115
C. vulgaris513 ± 59.84 ± 0.290.52 ± 0.008
E. angustifolium472 ± 45.68 ± 0.390.16 ± 0.008
E. vaginatum467 ± 614.00 ± 0.431.09 ± 0.008
P. sitchensis492 ± 417.02 ± 0.391.76 ± 0.015
S. auriculatum448 ± 37.27 ± 0.160.33 ± 0.008
Table 2.   Mass loss of five litter types after 6, 12 and 18 months, at three water tables (low, medium and high), expressed as a percentage of starting mass; mean ± SE, n=6
HarvestWater tableLitter type
C. vulgarisE. angustifoliumE. vaginatumP. sitchensisS. auriculatum
1Low84.8 ± 1.186.2 ± 0.485.4 ± 1.165.2 ± 1.0104 ± 3.6
Medium87.1 ± 1.189. 9 ± 1.687.4 ± 1.767.8 ± 2.0104 ± 3.1
High98.6 ± 2.197.9 ± 2.199.0 ± 2.981.8 ± 0.6111 ± 1.7
2Low69.6 ± 2.372.5 ± 3.870.4 ± 3.770.6 ± 2.6102 ± 3.3
Medium71.3 ± 4.471.6 ± 2.971.6 ± 1.670.6 ± 2.094.7 ± 2.5
High91.1 ± 1.093.5 ± 1.592.2 ± 1.477.8 ± 0.9108 ± 3.6
3Low60.4 ± 3.068.8 ± 4.666.4 ± 3.261.2 ± 2.1104 ± 3.8
Medium65.1 ± 1.268.1 ± 0.983.3 ± 1.762.3 ± 0.994.3 ± 2.5
High87.6 ± 1.583.9 ± 1.183.3 ± 1.775.5 ± 1.097.1 ± 1.3

Carbon dynamics

There was a striking pattern in change of C content between litter buried at different water tables that was consistent across most litter types except S. auriculatum (Fig. 1a–e): losses of C were greater from litter buried at the low and medium water tables than from that at the high water table. For C. vulgaris, E. angustifolium, E. vaginatum and P. sitchensis, 60–70% of the C remained in the litter by Harvest 3. Litter buried at the high water table contained over 85% of the original C, except for P. sitchensis (71%) by the third harvest. Water table was a significant factor at harvests (P<0.017; Table 3). Post hoc tests showed that there was no significant difference between C losses in litter at low and medium water tables but the changes in C in litter at the high water table were always significantly different from changes in C in the litter buried at the other two water tables (P<0.05; Table 4). Changes in the C content of C. vulgaris, E. angustifolium and E. vaginatum litter showed a similar pattern of progressive decline over the three harvests. In contrast changes in C content of P. sitchensis litter showed a more rapid rate of decline at the first harvest (75% of C remaining at Harvest 1), followed by slower rates of loss over the next two harvests (68% of C remaining at Harvest 2 and 65% remaining at Harvest 3; Fig. 1a–e). Sphagnum auriculatum litter increased its C content at the high water table over the first 52 weeks, after which it decreased. Overall, changes in C content of the different litter types were significantly different (P<0.001 at all harvests; Table 3). Change in C content of litter from P. sitchensis and from S. auriculatum were always significantly different from changes in C content of all the other litter types (Table 4); there was no significant difference between change in C content of litter from the two Eriophorum species, nor among litter from C. vulgaris and E. angustifolium and E. vaginatum. Variance partitioning showed that at Harvest 1, litter type was more influential than water table in determining changes in C content (64% and 21%, respectively; Table 3), but the difference was less at Harvest 2, when water table accounted for 30% of the variance and litter type 47%. At Harvest 3, water table only explained 22% of the variance and litter type also decreased to 39%, indicating a steady decline in the importance of litter type in determining changes in C over time. Overall, measured factors accounted for 83–88% of the variation shown. At Harvest 3, there was a significant interaction (P<0.001) between litter type and water table and this accounted for 18% of the variation. LSD tests (results not shown) revealed that there were no trends across litter type or water table.

Figure 1.

 Litter quality variables of five litter types (Calluna vulgaris, etc.) at three water tables over three harvests (6, 12 and 18 months). (a–e) Remaining C as a percentage of initial amount; (f–j) remaining N as a percentage of initial amount; and (k–o) remaining P as a percentage of initial amount. Symbols represent data from litter buried at low (•), medium (○) and high (▾) water table. n=6, means ± SE.

Table 3.   Variance partitioning (%) and significance (P value) from anova for amounts of C, N and P remaining in litter bags over three harvests (6, 12 and 18 months)
Percentage varianceP valuePercentage varianceP valuePercentage varianceP value
Harvest 1
 Block0.03 0.65 0.21 
 Water table20.90.0171.320.6620.20.035
 Residual3.18 5.60 4.67 
 Block/plot/set2.37 7.31 2.67 
 Litter type63.5<0.00151.8<0.00139.6<0.001
 Water table × litter type0.890.6611.00.0016.570.08
 Residual9.10 22.4 26.10 
Harvest 2
 Block0.27 0.79 2.15 
 Water table30.10.0133.420.3332.540.17
 Residual3.90 4.67 1.78 
 Block/plot/set2.29 5.33 5.49 
 Litter type47.3<0.00142.8<0.00167.7<0.001
 Water table × litter type2.940.12513.30.0036.92<0.001
 Residual13.24 29.7 13.4 
Harvest 3
 Block2.41 12.07 12.06 
 Water table21.70.0094.110.0728.70.006
 Residual2.28 1.49 2.33 
 Block/plot/set6.57 4.16 1.73 
 Litter type39.4<0.00142.57<0.00117.6<0.001
 Water table × litter type18.5<0.00113.04<0.00122.6<0.001
 Residual9.17 22.55 15.09 
Table 4.   Tukey's post hoc tests of significance following anova, among different litter types and water tables for C, N and P (as a percentage of original amount) remaining in litter bags over three harvests (6, 12 and 18 months)
  • *

    Significant at P<0.05; n=6.

  • NS, not significant.

Water tables
Litter types
 C. vulgaris/E. angustifoliumNSNSNS******
 C. vulgaris/E. vaginatumNSNS**NSNSNSNSNS
 C. vulgaris/P. sitchensis***NSNSNS**NS
 C. vulgaris/S. auriculatum***NS**NS*NS
 E. angustifolium/E. vaginatumNSNSNSNSNS****
 E. angustifolium/P. sitchensis*********
 E. angustifolium/S. auriculatum********NS
 E. vaginatum/P. sitchensis****NSNS***
 E. vaginatum/S. auriculatum******NS*NS
 P. sitchensis/S. auriculatum***NS*****

N dynamics

There were no consistent patterns among changes in N in the litter buried at different water tables and there were no significant differences among the three water table levels on N loss at any of the harvests (Table 3). While there was a significant interaction between water table and litter type at all harvests, LSD tests did not show any clear trends for particular litter types or water tables (results not shown). N accumulated in most litter types except C. vulgaris (Fig. 1f–j), resulting in increases in the amount of N from 2% to 58%; accumulation of N was the highest in E. angustifolium at all water tables. Across all litter types, increase in N content was the highest at the first harvest, after which N content generally decreased (N content of litter from P. sitchensis remained the same at the high water table between Harvests 1 and 2). At the third harvest, change in N ranged between a 40% increase in N content (E. angustifolium litter, low water table) and a 36% decrease in N content (S. auriculatum litter, medium and high water tables). At all three harvests, litter type had a significant effect (P<0.001; Table 3). Changes in N content in the combinations of litter type: C. vulgaris and E. angustifolium; E. angustifolium and P. sitchensis; E. angustifolium and S. auriculatum; E. vaginatum and S. auriculatum were significantly different at every harvest (Table 4). The only pair of litter types where change in N in the litter was never significantly different from each other was C. vulgaris and P. sitchensis. Although there were no significant differences among accumulation of N in the different litter types at any of the water tables (Table 3), there was a trend indicating that the percentage variation attributable to water table variation increased at each harvest, and the corresponding P values decreased to P=0.07 at Harvest 3 (Table 3). Differences in amount of N in the litter types accounted for 51% of the variance at Harvest 1, reduced to 43% at the second and third harvests. The interaction between litter type and water table accounted for 11–13% of the variance. Variance partitioning of block effects increased from around 0.7% at Harvests 1 and 2, to 12% at the final harvest; at the block/plot/set level, the range was between 4% and 7%. Overall, measured factors accounted for 66–76% of the variation shown.

P dynamics

There were no obvious patterns in the changes in P content of litter buried at different water tables (Fig. 1k–o), with both increases and decreases in P content. At Harvest 1, the change in P in litter buried at different water tables was significantly different between the low and high water tables only (Table 4). At Harvest 3, amounts of P in the litter buried at the low water table were significantly different from both the high and medium water tables. Overall the highest losses of P occurred from litter buried at the high water table (Fig. 1k–o). P accumulated in E. angustifolium and S. auriculatum litters at the low and medium water tables. In both cases there was more accumulation at the second harvest (except at the medium water table for E. angustifolium) followed by a decrease at Harvest 3. P content of litters was a significant variable at all harvests (Table 3); only the P content of C. vulgaris and E. vaginatum litters was never significantly different from each other (Table 4). Litter from P. sitchensis had the highest starting amount of P (Table 1) and showed the largest decrease across all three water tables (Fig. 1n). There was a significant interaction term of litter type and water table at the second and third harvests but LSD tests (data not shown) did not reveal any clear patterns or trends. Percentage variance in P content due to water table was similar at Harvests 1 and 3 (20% and 28%, respectively) but was only 3% at the second harvest. Percentage variance due to litter type accounted for 40% at Harvest 1, 68% at Harvest 2 and 18% at Harvest 3. Overall, measured factors accounted for 69–85% of the variance found. There was a strongly positive correlation between the absolute amounts of N and P in litter bags (rs=0.839; P<0.001). Twenty-eight per cent of the samples accumulated P, 58% accumulated N and 21% accumulated both N and P (data not shown).

Fungal community composition

The repeated fungal samples (Fig. 2) showed little variability due to PCR and gel differences. Axes 1 and 2 accounted for 20.6% and 18.6% of the variation, respectively. There were no residual differences among gels, when assessed using the PC scores (data not shown). Analysis of fungal community structure using PCoA showed that the first and second axes accounted for 21.1% and 9.6% of variance, respectively (Fig. 3). Figure 3a–h display the same data from the PCoA but split by litter type (Fig. 3a–d) and by water table (Fig. 3e–h). Change in fungal community structure over time is apparent in Fig. 3 where different PCoA coordinates at different harvests show changes in the fungal community composition. Proximity of data points represents degree of similarity among samples and size of the error bars shows the amount of variation among replicates associated either with water table or with litter type, on both the x and y axes. At time 0 (Fig. 3a) the fungal community composition associated with different litter types showed greater separation (and thus differences) compared with the later harvests, particularly Harvest 1 (Fig. 3b). Fungal community composition associated with E. angustifolium showed relatively large changes over time, while C. vulgaris and P. sitchensis showed little change between the second and third harvests (Fig. 3c–d). The analysis showed that there was very little separation between the fungal community composition found at different water tables at each harvest (Fig. 3e–h), especially at time 0 (Fig 3e), but that overall, the fungal community composition changed at each harvest. At Harvests 1–3 (Fig. 3f–h), fungal community composition associated with low and medium water tables are more similar than those at the high water table; this effect is less apparent at the third harvest.

Figure 2.

 Principal coordinate analysis showing PCR and gel variability in repeated samples for each litter type and the standard (Hyperladder); for litter types n=5; for standard n=6; mean ± SE.

Figure 3.

 Principal coordinate analysis of DNA extracted from fungi colonizing litter from five different plant species (Calluna vulgaris, Eriophorum angustifolium, Eriophorum vaginatum, Picea sitchensis and Sphagnum auriculatum) buried at three water tables (low, medium and high) and harvested after 6, 12 and 18 months. Symbols represent data from litter of C. vulgaris (○), E. angustifolium (▿), E. vaginatum (□), P. sitchensis (⋄) and S. auriculatum (▵) at low (•), medium (▾) and high (▪) water table. Axis 1 accounted for 21% of the variation and axis 2 accounted for 9.6%. For time 0, n=3; for Harvests 1–3, n=6.

Analysis of presence/absence band patterns for each harvest using distlm showed that differences in fungal community structure due to litter type and water table were significant at each harvest (P<0.001; Table 5). The interaction between litter type and water table was not significant for the first two harvests, but was highly significant (P<0.001) at the third harvest. Block effects were significant at each harvest (P<0.001). Initially, differences in fungal community structure due to different litter types accounted for 24% of the variation seen. At each harvest, litter type accounted for 8%, 12% and 13% of the variance, respectively. Subsequent differences in fungal community composition caused by burying litter at different water tables initially accounted for 0.59% of the variation (time 0). Differences in fungal community composition due to water table accounted for c. 7% at each harvest. At time 0, interaction between differences due to water table and litter type accounted for 13% of the variance, this decreased to c. 8% at Harvests 1 and 2, while at the final harvest interaction effects accounted for 14%. Block effects accounted for c. 12% of the variance at each harvest. In total, measured factors only accounted for 38–47% of the variation found.

Table 5.   Variance partitioning (%) and significance (P values) of fungal community composition from DGGE analysis of fungi colonizing litter bags over three harvests (6, 12 and 18 months), with five litter types and three water tables using distlm
Time 0Percentage varianceP value
  1. Means ± SE; for time 0, n=3, for Harvests 1–3, n=6.

Litter type24.30.0001
Water table0.590.9
Litter type × water table12.90.8718
Harvest 1
 Litter type8.480.0015
 Water table7.650.0001
 Litter type × water table8.450.401
 Block effects11.80.0008
Harvest 2
 Litter type12.00.0001
 Water table7.360.0001
 Litter type × water table7.140.63
 Block effects13.50.0001
Harvest 3
 Litter type12.920.0001
 Water table7.520.0001
 Litter type × water table14.50.0001
 Block effects11.60.0002

CCA plots for Harvest 1 showed water table separating along axis 2 (Fig. 4a) with most of the data from the high water table (white triangles) separated from the low and medium water tables, whereas there was less separation between the low and medium water tables. The eigenvalues for %C and %P showed that these were the main differences separating fungal community composition within the variables measured. There was more overlap between the low and medium water tables, where %P appeared to be more important, and the high water table where %C was more important. The different litter types separated along axis 1 (Fig. 4d) and showed overlaps between S. auriculatum and C. vulgaris and between E. vaginatum and E. angustifolium. The eigenvalues showed that there was a slight association between %C and S. auriculatum and between %N and %P and the two Eriophorum spp. Picea sitchensis (black squares) also showed a weak association with %P. All three litter quality variables were significant (P<0.05; Table 6), both independently and following forward selection, but the total variance in fungal community structure explained by the %C, N and P of litter was only 9%.

Figure 4.

 Canonical correspondence analysis of fungal community composition from five litter types buried at three water tables. Symbols represent data from litter buried at low (○), medium (▾) and high (▵) water table and litter of Calluna vulgaris (○), Eriophorum angustifolium (▾), Eriophorum vaginatum (▵), Picea sitchensis (▪) and Sphagnum auriculatum (•). Eigenvalues show change in C, N and P at each harvest, expressed as a percentage of the original amounts in the litter. For (a) and (d), axis 1 accounted for 5.0% of the variation and axis 2 for 2.4%; for (b) and (e), axis 1 accounted for 15.0% of the variation and axis 2 for 5.8%; for (c) and (f), axis 1 accounted for 5.9% of the variation and axis 2 for 2.9%. See Table 5 for percentage variation explained by each litter quality variable at each harvest.

Table 6.   Results from canonical correspondence analysis showing percentage variance in fungal community composition of litter bags explained by environmental variables (%C, %N and %P) over three harvests (6, 12 and 18 months)
 Percentage variance
explained by individual
Percentage variance
explained after forward
  1. *P<0.05 ; **P<0.01 ; ***P<0.001.
    NS, not significant.

Harvest 1
 Total variance explained9.0
Harvest 2
 %P5.00 NS6.0*
 Total variance explained24.0
Harvest 3
 %C2.0 NS2.0 NS
 %P3.0 NS4.0*
 Total variance explained11.0

CCA plots for Harvest 2 showed that the low (white circles) and medium (black triangles) water tables (Fig. 4b) were separated from the high water table (white triangles). As with Harvest 1, the eigenvalue for %C was associated with high water table; %N and P were more associated with low and medium water tables. Splitting the data by litter types showed a separation along axis 1, with clear separation of points representing S. auriculatum (black circles) and all other litter types (Fig. 4e). Percentage C and %N, accounted for 11% and 7% of the variation, respectively (Table 6). Again, the eigenvalues suggest that S. auriculatum litter (black circles) was associated with %C and that E. angustifolium litter (black triangles) was associated with %P. The three litter quality variables explained 24% of the variation.

At Harvest 3, the data showed less separation with greater overlap suggesting that there were no clear associations between litter quality variables and fungal community structure found on litter at different water tables (Fig. 4c) and much less separation. There was less separation due to litter type (Fig. 4f) with %C, N and P only accounting for 11% of the variance (Table 6). Percentage C was weakly associated with S. auriculatum and %N and %P with the fungal community structure found in E. angustifolium litter.


We found that litter type had a highly significant effect on fungal community composition throughout the experiment and that this was slightly more important than water table (Table 5). Variance partitioning showed litter type accounted for at least twice as much of the variance in the time 0 samples compared with its importance during decomposition at the later harvests. As plant material senesces, it is possible that endophytic species will increase in activity, competing directly with colonizing saprophytic species; however, the observed decrease in variance in the fungal community composition that was attributed to litter type between the start of the experiment and the first harvest may also suggest a decline of the endophytic fungal population and/or substitution with a more generalist population of early saprotrophs. Similarly, the presence of endophytes in grass species has been found to reduce rates of decomposition (Lemons et al., 2005). From Harvests 1 to 3, the percentage variance due to fungal community composition associated with different litter types increased from 8% to 12% showing that as decomposition proceeds, the differences between fungal community composition of the different litter types became greater. This may be due to fungal communities moving from r to K selection; this has been observed in bacterial communities (Noll et al., 2005), and litter decomposition communities (e.g. Waldrop & Firestone, 2004); relative amounts of such compounds are likely to vary among the different litter types (Bragazza et al., 2007). Variance partitioning indicated that there were significant differences in fungal community composition due to water table but this remained constant (c. 8%) throughout the experiment. The lack of a significant effect of water table at time 0 in comparison with the samples from the three subsequent harvests was likely because this material was rewetted with distilled water for 48 h to normalize conditions for DNA extraction, rather than having been placed in the experimental site as a water table treatment. Differences in fungal community composition due to interactive effects between litter type and water table varied from 7% to 14%, but were only significant at the final harvest. LSD tests confirmed the importance of litter type and the high water table in determining decomposition; S. auriculatum was the only litter type that showed significant differences in change in C depending on the water table at which it was buried (data not shown).

Despite the overall similarity among fungal community structure associated with different litter types, there were clear changes over time according to the litter type (Fig. 3a–d) and water table (Fig. 3e–h) at which the litter was buried, suggesting that the fungal assemblages undergo a successional shift. This is most apparent at Harvest 2 where fungal community composition of the different litter types shows greater separation than those at time 0 and Harvests 1 and 3. Other studies present a mixed picture on this issue. For example, Thormann et al. (2003) used culturing techniques on simple sugars as C sources to investigate microfungi colonizing a number of different litter types in a peatland in Canada. They only observed a clear succession of fungal community composition for two out of the five litter types, leading them to conclude that overall the classic pattern of fungal succession did not occur in peatlands. Instead, they identified a suite of generalist microfungi that coexisted and decomposed the litter simultaneously. They did not identify species able to decompose complex compounds such as lignin and suggest this could be due to the short timescale of the experiment (2 years), although this could also be due to the limitations of the culturing methods used. Artz et al. (2007), using molecular techniques, observed changes in fungi associated with vegetation succession in regenerating cutover peat where harvesting finished different lengths of time ago, including a large increase in fungal sequences from basidiomycetes in areas abandoned longer ago. In our study, 53–63% of the variation remained unexplained by litter type and water table (Table 5). Variation due to the block structure of the experiment was similar to variation due to litter type and higher than that due to water table. One explanation for this could be the high heterogeneity of the residual, bare, peat, which was reported to show considerable variation in the fungal community composition in different parts of the experimental site (Artz et al., 2007) and thus might affect colonization of litter bags by fungi.

Few studies have used molecular fingerprinting techniques such as DGGE to compare fungal community structure in peatlands. To our knowledge, this is the first time this method has been used to relate fungal community composition of decomposing litter on a cutover peatland with changes in litter quality, differences in depth of water table and the interactive effects of these variables. We found that the C, N and P concentrations were significant explanatory variables of fungal community structure (except for C at the third harvest), although the majority of the variation in fungal community composition (76–91%) remained unexplained. Other possible factors affecting fungal community composition include water chemistry, pH and physical variables such as temperature. In a similar experiment where fungi were cultured from decomposing litter from natural peatlands, Thormann et al. (2004) found that none of these factors was significant. However, in an experiment investigating microfungal succession on litter in undamaged peatlands, Thormann et al. (2003) reported that pH was a significant factor influencing the fungi associated with mid to late stages of succession in Sphagnum fuscum (Schimp) Klinggr. Peat is highly heterogenic and it is possible that differences in the quality and structure of peat affects the fungal communities present in the peat and thus able to colonize litter in any particular location. Very little mixing of peat occurs from the activity of mesofauna. Although peat does not support earthworms, there can be large numbers of enchytraeid worms in some organic soils (Briones et al., 1998). These are responsible for mixing organic material and may fulfil a similar role to earthworms in mineral soils. However, the importance of enchytraeids on cutover sites is unknown and the old, deep peat of cutover peatlands exposed at the surface is unlikely to provide suitable conditions (Laiho, 2006). The subsequent lack of mixing of organic material could result in differences in the peat determined largely by the plant species present at particular locations and times.

The quality and dynamics of C sources in the litter are likely to be an important regulator of microbial community structure. For example, it has been shown that the structure of microbial communities from woodland and grassland can be separated based on use of complex substrates rather than use of single C sources (Waldrop & Firestone, 2004). Additions of dissolved organic C leached from forest litter have also been shown to lead to profound shifts in the abundance of certain members of the soil microbial community (Cleveland et al., 2007). Previous experiments analysing the composition of fungal communities on decomposing litter in natural peatlands have also found relationships between these variables. For example, Thormann et al. (2004) found a clear separation between fungi isolated and cultured from different litter types and correlations with C : N quotient, total P and total N, but they found litter quality variables explained far more variation than our data (axes 1 and 2 explained 96% of the variation, compared with 7.4%, 21% and 8.8% in our study). This is interesting, given that culturing is well known to reflect only a small fraction of the total fungal population and is likely, therefore, to have represented an incomplete picture of fungal community composition in peatlands. Nevertheless, their study suggested that culturing methods were able to identify enough of the functionally active fungi to explain the majority of variation in litter quality over time. It is possible that these associations are obscured when using molecular techniques where all fungal DNA is amplified regardless of its activity or relative abundance (Artz et al., 2007), although direct comparisons between the diversity of culturable and DNA-based fungal populations from identical samples would be required to test this thoroughly. In addition, fungi present when litter was buried may not be active but could persist through part or all of the experiment, and be recorded along with new microorganisms colonizing the litter at each subsequent harvest and this could account for small overall changes in fungal community composition over time (Fig. 3). Alternatively, the fungi present on the litter at the time of burial may be responsible for initial losses of labile C. In addition, there is an assumption of competitive hybridization in the presence of excess primers (Bridge & Spooner, 2001; Anderson & Cairney, 2004) but it is possible that PCR also amplifies DNA present only at very low frequency. While this type of PCR bias has been identified as a problem in some situations (Dahllöf, 2002), the high similarity of repeated samples (Fig. 2) suggests that this was unlikely to be a major issue in our experiment.

Overall the losses of C from each litter type fitted the expected pattern that more C would be lost from the low water table than high water table treatment. Although there were significant differences between changes in litter C at the high water table and both low and medium water tables, surprisingly, there was no difference in the change in litter C between the low and medium water tables, leading us to reject our third hypothesis. This lack of difference could be because, regardless of the distance to the water table, the top few centimetres of peat desiccate quickly during dry weather and resaturate when it rains, so that distance to the water table has little influence once this is greater than the depth at which the litter bags were buried (c. 6 cm). In such situations, water content of the peat may be a more appropriate measure than distance to the water table.

Contrary to our third hypothesis that little C would be lost from the litter at the high water table treatment (where litter was submerged), 9–29% of the C had been lost from litter by the third harvest (Fig. 1a–e). Loss of C from C. vulgaris, Eriophorum spp. and P. sitchensis litters at the high water table was surprising given that fungi are not generally considered to be important decomposers in submerged conditions (Bossio & Scow, 1998), especially when the water is also acidic (Deacon, 1997). These litter samples clearly supported different fungal communities while submerged and these communities changed over time (Fig. 3), which suggests that loss of C was likely to be due to active fungal growth, although we cannot exclude competitive C utilization by other microorganisms.

Variance partitioning showed that the relative importance of litter type and water table in determining changes in C content of litter varied over the three harvests (Table 3), with litter type becoming less important over time. This might indicate that in the first 12 months the litter types had different labile C compounds that were utilized by microorganisms at different rates, while in the final 6 months, the C compounds remaining were similar, or at least similarly recalcitrant. Certainly there is evidence that the amount of labile C in litter is a key regulator of initial decay rates (e.g. Gessner, 1991; Gillon et al., 1994), just as the type of C present or remaining is likely to be a good indicator for successional fungal population structure (e.g. Osono, 2005). This may also account for the significant interaction at the third harvest, when an increasing proportion of recalcitrant C could result in litter type responding differently at the three water tables. It is not clear why the effects of water table should increase at the second harvest, accounting for 30% of the variance, compared with c. 20% for Harvests 1 and 3. Seasonality is known to have a strong effect on soil microbial communities (Williams & Rice, 2007) and this may provide some explanation of the results because Harvests 1 and 3 took place in the summer while the second harvest was in winter.

There were very few changes in C content of S. auriculatum litter at any of the water tables; this is consistent with the role of Sphagnum spp. as the principal peat-forming plants and confirms the findings of other studies (Aerts et al., 1999; Bragazza et al., 2007). Different Sphagnum spp. vary in their rates of mass loss, and comparison of our data (5% losses to 2% gains after 1 year) with other studies of Sphagnum decomposition shows that our results lie at the lower end of the range of variation reported (0.1–20% losses; Johnson & Damman, 1991). The slight increase in mass in some of the litterbags containing S. auriculatum (Table 2) likely reflects ingress of peat particles that were not removed by subsequent washing of the samples.

Eriophorum angustifolium and S. auriculatum litter, which had the lowest P concentrations at the start of the experiment, accumulated P at the low and medium water tables whereas P tended to be lost from the litter at the high water table. As with N, accumulation of P is not uncommon during decomposition although increases in P of the scale we recorded are rare. For example, similar levels of P accumulation occurred in Carex species in low productivity mesotrophic fens in the Netherlands (Aerts & deCaluwe, 1997), while increases in P of 133–188% have been recorded in Pinus banksiana (Lamb) needles over 12–24 months (MacLean & Wein, 1978). They ascribe these increases to low soluble P concentrations in soil resulting in microorganisms strongly binding P. In tundra, microorganisms act as strong sinks for P especially when plants are not present (Jonasson et al., 1999), as was the case in this experiment. In nutrient-poor fens in the Netherlands, accumulation of P is associated more with Sphagnum- than Carex-dominated sites, with up to 140% immobilization recorded for Sphagnum papillosum Lindberg after 6 weeks (Scheffer et al., 2001).

When fungi are limited by particular nutrients, they are able to translocate them from the environment (Laiho, 2006). For each set of litter bags, the combined losses of P were always greater than the combined increases (data not shown) so it is possible that mineralized P from adjacent litter bags could have acted as the source of P. However, movement of N among decomposing leaves is regulated by source strength rather than sink strength (Schimel & Hättenschwiler, 2007); if the same is true for P, the changes in P that we observed are perhaps more related to losses from high P litters (particularly P. sitchensis) rather than accumulation of P in the litter with low P concentrations (E. angustifolium).

All litter types accumulated N to a greater or lesser extent, leading us to accept our fourth hypothesis but there was no clear pattern of accumulation at different water tables. Accumulation of N is a commonly observed phenomenon in the early stages of decomposition and occurs particularly where C : N quotients are higher than the theoretical optimum of 30 (Deacon, 1997). As C : N quotients reduce, accumulation reduces until net mineralization starts to occur. There was a trend towards mineralization by the third harvest (except with E. angustifolium, which had the lowest concentration of N at the start of the experiment). However, we did not find C : N quotients were good predictors of the shift from accumulation to loss of N (data not shown), in contrast to other studies (e.g. Scheffer et al., 2001). This suggests that in our experiment, the form of C present, for example as lignin or phenolics, may be more important in determining decomposition rates (Heal et al., 1997). In contrast to P, there was no obvious pattern between the amounts of N lost from or accumulated in different litters and it is therefore unlikely that translocation of N from one litter type to the other was a major factor.

The strong positive correlation between amounts of N and P in litter bags suggests that both N and P were transported simultaneously by the same mechanism. This in turn suggests that the source of the P accumulating in litter was not adjacent litter bags. There were no correlations between percentage change in N and P and corresponding PCoA scores (data not shown).

In many cases, we observed significant interactive effects between water table and plant litter type. Although no clear trends were obvious, these results confirmed that S. auriculatum litter behaves very differently from the other litters in terms of change in C, N and P. Significant interactions also confirmed that there are few differences between decomposition at the low and medium water tables (data not shown).

This work has demonstrated the significance of water table in determining rates of decomposition on a cutover peatland. This is an important finding because water table manipulations are a key management strategy for restoring abandoned peatlands. In addition, the hydrological conditions of peatlands are predicted to change alongside global climate. The high water tables lead to a reduction in rates of decomposition, confirming the importance of this type of management on cutover sites in order to reduce C lost through microbial activity. When water tables are high, the species mixture of native colonizing plants is likely to be less important as decomposition proceeds at a slower rate. Although S. auriculatum showed the lowest rates of decomposition, this genus is restricted to areas where water tables are close to the surface. However, for water table to be a significant regulator of decomposition, our results suggest that it needs to be very near the surface – we found no difference in decomposition between litter when the water table was on average 9 or 18 cm below the peat surface. Our data provide striking evidence that the depth of water table and plant litter type strongly affect the composition of the fungal community and the turnover of the key elements C, N and P but we still need to gain a much fuller understanding of the fungi found in these habitats and the different roles they play in decomposition of plant litter.


C.J.T. was supported by a BBSRC PhD studentship and the Macaulay Development Trust. R.R.E.A. was supported by the Scottish Executive and D.J. receives funding through an NERC Advanced Fellowship. We thank George Watson for access to Middlemuir Moss and details of the harvesting history; and are grateful to Janis Brodie, Clare Cameron, Angela Fraser, Dot Mackinnon, Pamela Parkin, Eileen Reid, Nadine Thomas and Duncan White for assistance with analytical techniques, and Marti Jane Anderson (University of Auckland, New Zealand), Betty Duff and Jackie Potts (Biomathematics and Statistics Scotland) for advice on statistical analysis.