Introduction

The key processes in terrestrial biogeochemistry—primary production, decomposition, and nutrient cycling—are largely controlled by the interaction of cycles of biologically important elements. These cycles are generally closely coupled due to basic constraints on the elemental composition of biological macromolecules leading to an emergent stoichiometry of processes from the molecular to ecosystem level (Sterner and Elser 2002). In many ecosystems, the interaction between C and N cycles is of particular importance. For example, available nitrogen often limits gross primary production (LeBauer and Treseder 2008), can alter patterns of C allocation in forests (Vicca et al. 2012), and may influence the rate of soil organic matter turnover and thus control soil carbon storage (Janssens et al. 2010; Knicker 2011). The current research effort to understand terrestrial carbon cycling in the context of climate change therefore requires understanding of the interactions between C and N cycles, as well as the response of these linkages to global changes such as climate warming (Finzi et al. 2011).

Decomposition of organic matter in soils is an important point of connection between terrestrial C and N cycles. The rate of breakdown of polymeric forms of C determines the magnitude of carbon storage and emissions of greenhouse gases, and has feedback effects on primary production by making organic and mineral N available for re-uptake by plants. At large scales decomposition is controlled by environmental conditions and, in some contexts, the chemical composition of the organic matter inputs (Aerts 1997; but see Schmidt et al. 2011). However, at the molecular scale the proximate drivers are exoenzymes produced by soil microbes, which catalyse the hydrolytic and oxidative reactions that breakdown polymeric organic substances (Nannipieri et al. 2002). Bacteria and fungi produce a wide range of enzymes, each specialized for the breakdown of specific classes of molecules e.g. alpha-glucosidase (starch), cellulase (cellulose), phenol oxidases (complex polymeric substances such as lignin) and proteases and peptidases (polypeptides) (Skujinš 1976). The production of these enzymes is regulated by both the availability of the substrate, and the microbial demand for the corresponding product (Allison and Vitousek 2005; Allison et al. 2011). Moreover, the enzymes themselves are N-rich and energetically costly to produce. The relative levels of production and activity of different groups of soil extracellular enzymes therefore reflects the physiological constraints and economic trade-offs within the microbial community (Schimel et al. 2007; Allison et al. 2010).

Recent developments in decomposition modelling have incorporated the dynamics and physiology of microbial biomass and the production and activity of extracellular enzymes (e.g. Schimel and Weintraub 2003; Moorhead and Sinsabaugh 2006; Allison et al. 2010). These models reveal the potential for subtle interactions between environmental conditions, substrate supply and decomposition rates, and have shed light on previously puzzling phenomena such as the observation that N addition can suppress decomposition rates (Craine et al. 2007). Moreover, there is an ongoing discussion about how best to use enzymes to improve the realism of large-scale earth system models (Todd-Brown et al. 2012). However, a common assumption of enzyme-based models is that the dynamics of C- and N-acquiring enzymes can be modelled together, that is, that they show similar responses to environmental variability in terms of levels of enzyme production. Whether this assumption is valid, particularly in terms of the similarity of responses to climate change phenomena, is an important question with implications for the realism of model predictions of soil organic matter decomposition.

There have been several recent studies examining the effects of experimental warming on soil enzyme activities (reviewed in Henry 2012). These studies usually measure the potential activities (pool sizes) of various enzymes, but shifts in enzyme kinetics (temperature sensitivity of maximum reaction rate and/or substrate affinity) may also play an important role in in situ soil responses to warming (Wallenstein et al. 2011). It has been shown that temperature sensitivity (Q10) of maximum enzyme reaction rates (Vmax) can vary with season and/or soil type (Trasar-Cepeda et al. 2007; Wallenstein et al. 2009), but how this parameter is affected by long term soil warming is unknown. Furthermore, studies of warming effects on soil enzymes tend to examine each enzyme separately, whereas an integrative approach that explicitly compares the dynamics of a range enzymes involved in both the C and N cycles may help to elucidate the extent to which they are subject to the same environmental controls.

In this study, we investigated the response of a range of hydrolytic soil enzymes to experimental climate manipulation in a long-term experiment in northern Sweden. We measured four enzymes each from two functional groups. We measured glycosidases as indicators of soil C cycling. These enzymes catalyze hydrolytic degradation of polymeric carbohydrates such as cellulose and starch into simpler sugars. To investigate organic N cycling we assayed four amino-peptidases-enzymes which catalyse the hydrolysis of peptide bonds in polypeptides. It is important to note that peptidase activity releases assimilable C as well as N, which complicates its interpretation as an indicator of N-cycle dynamics (Sinsabaugh and Foreman 2003). For the purposes of our interpretation we make the assumption that peptidase activity is related to N supply and/or microbial demand in a meaningful sense (Allison and Vitousek 2005; Sinsabaugh et al. 2008), but we cannot strictly rule out a simultaneous relationship with C dynamics.

Our study particularly focused on testing whether or not enzymes from the two functional groups responded similarly to the climate treatments. Furthermore, we used samples from the beginning and end of the summer growing season, as both potential activity and temperature sensitivity can show significant seasonal variation (Bonnett et al. 2006; Wallenstein et al. 2009; Bell et al. 2010), and therefore any patterns of response to warming may vary over time. Our experimental site is located in a sub-arctic peatland, a biome of special significance in the global carbon cycle due to its status as a long term carbon sink (Limpens et al. 2008) in a region that is particularly vulnerable to climate warming (ACIA 2004). The links between C and N cycles in this system are also important: N availability strongly limits C cycle processes (Rosswall and Granhall 1980), and components of both cycles have been shown to respond strongly to climate warming (Dorrepaal et al. 2009; Weedon et al. 2012).

Our study addresses the following questions:

  1. 1)

    How do the potential activities and temperature sensitivities of soil glycosidases and peptidases respond to experimental climate manipulation?

  2. 2)

    Do the enzyme activities and temperature sensitivities vary between the beginning and end of the growing season?

  3. 3)

    Do soil glycosidases and peptidases show similar responses (i.e. same direction) to experimental climate manipulation and/or across the season?

Materials and methods

Study site and sampling

Soil samples for enzyme analysis were collected from a long-term climate manipulation experiment established on a blanket bog about 800 m from the Abisko Scientific Research Station in Abisko, sub-arctic Sweden (68°21′N, 18°49′E, alt. 340 m) (Dorrepaal et al. 2004). The gently sloping site is dominated by Sphagnum sp. mosses (predominantly Sphagnum fuscum), with a sparse cover of dwarf shrubs Empetrum hermaphroditum, Rubus chaemomorus, Betula nana and Vaccinium uliginosum. The climate is continental sub-arctic with mean annual rainfall of 352 mm and mean monthly temperatures in January and July of −9.7 and 12.3 °C respectively (meteorological data 1999–2008, Abisko Scientific Research Station). Since 2000 a climate manipulation experiment has been maintained to investigate the effects of summer and spring warming and winter snow accumulation on a range of soil and vegetation processes (Dorrepaal et al. 2004; Aerts et al. 2007; Dorrepaal et al. 2009; Keuper et al. 2011; Aerts et al. 2012). The experiment consists of factorial combinations of summer treatments (ambient or warming), and winter/spring treatments (ambient and snow addition + spring warming) randomly assigned to plots in 5 contiguous blocks parallel to the bog slope (i.e. 5 plots per 4 treatments = 20 plots total). Treatments are applied using open top chambers (OTCs): hexagonal, transparent polycarbonate structures 50 cm high, and with diameter 1.6–1.8 m at the top and 2.2–2.5 m at the base. For summer warning plots the OTCs are in place every year from the beginning of June until the end of September. For snow-addition + spring warming plots the OTCs are also in place for the remainder of the year (Dorrepaal et al. 2004). The OTCs increase average daily mean air temperature by 0.3–1.0 °C in spring (April–June) and by 0.2–0.9 °C in summer (June–October). Effects on soil temperature vary with depth and from plot to plot: but typically passive air warming also increases average soil temperature at 10 cm depth by 0.6–1.0 °C. During winter, there is a passive accumulation of snow leading to approximately a doubling of the snow layer thickness and an increase of winter average soil temperature of 0.5–2.2 °C (Dorrepaal et al. 2004; Dorrepaal et al. 2009). In spring the melting residue of this increased snow cover can also locally counteract the spring warming effect. Soil temperature in a subset of plots at 5, 10 and 20 cm depth is logged year-round at two-hourly intervals using methodology detailed in Dorrepaal et al. (2004).

Soil for enzyme analyses was sampled on June 4, 2009 and August 16, 2009. These dates were chosen to roughly coincide with the beginning of the vegetation growing season (June) and the latter part of the growing season, prior to vascular plant senescence (August). Soil cores of 5 cm diameter were taken to 15 cm depth near the centre of each plot. At this depth the soil is primarily partially decomposed Sphagnum material (i.e. Oi horizon) with organic matter content >95 % and % C approximately 45 % (±0.64 standard error; E. Dorrepaal, pers.comm). Within 12 h of sampling, samples were mixed and all coarse roots and live moss removed. Subsamples were taken for immediate measurement of soil water content. The remaining material was then immediately stored at −20 °C and transported to Amsterdam for further analyses. Although performing assays on freshly sampled material is preferable, logistic constraints made this impractical, and there is evidence that freezing at −20 °C does not significantly alter measured enzyme activities relative to cool storage (DeForest 2009). It should be noted, however, that although relative treatment effects were most likely not influenced by storage, absolute values of potential enzyme activity (see “Results” section below) may deviate from those that would have been obtained from freshly sampled soil.

Enzyme assays

Potential enzyme activity assays were conducted as described in Weedon et al. (2012) using the protocol of Steinweg and McMahon (http://enzymes.nrel.colostate.edu/). Model substrates labeled with the fluorophores 7-amino-4-methylcoumarin (MUC) or methylumbelliferone (MUB) were used to quantify the relative pool sizes (i.e. activity under saturating substrate concentrations) of enzymes responsible for the hydrolysis of four peptide and four carbohydrate substrates. The specific model substrates used were, for peptidase activity: l-leucine-7-amido-4-MUC (henceforth, Leucine), l-alanine-7-amido-4-MUC (Alanine), l-lysine-alanine-7-amido-4-MUC (Lys-Ala), and l-alanine-alanine-phenylalanine-7-amido-4-MUC (AAP); and for glycosidase activity: 4-MUB-B-d-xylopyranoside (xylase), 4-MUB-β-glucopyranoside (beta), 4-MUB-α-glucopyranoside (alpha) and 4-MUB-β-cellobioside (cellulase) (all substrates supplied by Sigma-Aldrich). The protocol includes construction of seven-point calibration curves for each individual slurry preparation and thus provides better correction for non-linear fluorescence quenching dynamics than the usual single-point correction.

We followed the protocol cited above using slurries created by homogenizing 4 g fresh weight of soil in 90 mL 0.5 M sodium acetate buffer (pH 5). Incubations were conducted at two temperatures (4 °C for 22 h or 20 °C for 4 h) to allow calculation of the temperature sensitivity of the hydrolysis of each of the model substrates. Temperature sensitivity estimates can depend on the choice of temperature range (Chapman and Thurlow 1998). In this study temperatures were chosen to reflect the natural range of soil temperatures at the site during the study period (see Fig. 1). End-point fluorometric measurements were made on a Spectramax Gemini XS microplate fluorometer (Molecular Devices, Sunnyvale, USA) with excitation wavelength of 365 nm and emission detection at 450 nm. All measurements were converted to nanomoles per gram dry weight per hour for statistical analysis.

Fig. 1
figure 1

Daily averages for minimum and maximum (dashed lines) and average soil temperature (solid lines) at 5 cm depth in summer warmed (black) and ambient (grey) plots from May to October 2009. Arrows indicate the dates of soil sampling for enzyme assays (4 June and 16 August 2009)

Full size image

Temperature sensitivities for enzyme activity were calculated for each sample × enzyme combination using the formula:

$$Q_{10} = e^{{\left( {10 \times \left( {\frac{{\ln A_{20} - \ln A_{4} }}{16}} \right)} \right)}}$$

where Q10 is the factor increase in enzyme activity with a 10 degrees increase in temperature, and A20 and A4 are the measured enzyme potential activities at 20 and 4 °C, respectively.

Statistical analysis

To test for the (possibly interactive) effects of climate manipulation treatments and sampling time on potential enzyme activities and temperature sensitivities, we fitted mixed-effects models for each enzyme separately (and for enzyme activities, separately for measurements at 4 and 20 °C). Sampling month, summer treatment and winter/spring treatment (and all two- and three-way interactions thereof) were included as fixed effects and plot ID modelled as a random plot-level intercept to account for non-independence of repeated samples taken from the same plot. The parameter estimates from these analyses suggested contrasting patterns of seasonal responses across the different enzymes, which we subsequently tested in a larger, four-way factorial mixed model combining all measurements (centred and scaled per enzyme, due to the large range in absolute values, and to focus on patterns in relative responses within each enzyme to treatment and sampling time). In this model we treated enzyme as a fixed factor, and all other factors as described for the three-way models above.

Results

The summer warming treatment led to soil warming over the period of the sampling (Fig. 1). The daily average temperature at 5 cm depth in summer warmed plots was, on average, 1.1 °C higher than in ambient plots between May 6 (1 month before first sampling) and August 16 (second sampling). Over the same period, daily maxima were on average 2.3 °C higher in warmed plots, with no difference in daily minima (n = 8 monitored plots for each treatment). The effects of the “spring warming + snow addition” treatment on soil temperatures in the 1 month preceding sampling was not statistically significant from ambient plots (ANOVA, P > 0.05)—due to a high variability in soil temperatures between these plots (data not shown), most likely due to patchily distributed residual snow.

Soil moisture content was not significantly affected by climate treatments but showed a significant (repeated measures ANOVA, P < 0.05) decline from June to August. The decline was, however, small in magnitude: from 88.0 ± 0.3 % water by weight (mean ± SE) in June, to 86.9 ± 0.5 % in August.

There was no significant effect of climate manipulation treatment on either the potential activity at 4 or 20 °C, or on temperature sensitivity of any of the eight measured enzymes (Tables 1 and 2). There was, however, a significant effect of sampling time on potential activity at 20 °C and this effect was enzyme dependent (Table 2, enzyme × month interaction P < 0.05). When tested individually, 6 out of the 8 enzymes, showed this seasonal effect. All peptidases had lower potential activity in August compared to June. Of the four glycosidases, only alpha-glucosidase had a lower activity in August (Fig. 2). Beta-glucosidase showed the opposite pattern, having significantly higher potential activity in August, and cellulase and xylase potential activity did not significantly change between the sampling events (Fig. 2). Almost identical patterns were seen in the measurements made at 4 °C, with the exception that lysine-alanine-peptidase showed no significant seasonal shift, while cellulase and xylase potential activity also decreased in potential activity between June and August (data not shown).

Table 1 Mean potential enzyme activity (nmol/g DW peat/hour) and its standard error (in parentheses) as measured by fluorometric enzyme assays over 4 h at 20 °C
Full size table
Table 2 ANOVA table for the fixed factors in linear mixed models for temperature sensitivity (Q10 values) and potential activity (at 20 °C)
Full size table
Fig. 2
figure 2

Mean (±95 % confidence interval) seasonal shifts (June to August) in potential enzyme activity measured at 20 °C. Negative values indicate reduction in potential activity over the growing season. Data for each enzyme are centred and scaled by standard deviation, to allow comparison across enzyme types, unscaled values are given in Table 1

Full size image

Similarly, apparent temperature sensitivity differed on an enzyme specific-basis between the two sampling moments (Table 2, enzyme × month interaction P < 0.05). All four peptidases showed a decrease in temperature sensitivity from June to August, whereas the glycosidases showed the opposite pattern (Fig. 3). The mean values of Q10 for the glycosidases were 1.8 (±0.2, standard error of mean) in June and 2.9 (±0.3) in August, whereas for the peptidases the values were 2.6 (±0.1) in June and 1.9 (±0.1) in August.

Fig. 3
figure 3

Seasonal shifts in temperature sensitivities (Q10) for four peptidase and four glycosidase enzymes, based on measurements at 4 and 20 °C. Points and bars represent means and 95 % confidence intervals respectively (n = 5)

Full size image

Discussion

The experimental climate manipulations imposed in our study had no effect on either the potential activities or temperature sensitivities of any of the 8 enzymes we assayed, a result that contrasts with the clear effects of the same treatments on soil respiration, nitrogen cycling and vegetation growth (Dorrepaal et al. 2009; Keuper et al. 2011; Weedon et al. 2012). On the other hand, we observed striking seasonal patterns in both potential activity and temperature sensitivity of almost all of the enzymes. These results have implications for our understanding of seasonal nutrient dynamics in this system, and raise issues about the incorporation of soil enzyme dynamics into models of soil processes and their responses to global change.

Lack of warming effect on soil enzymes

Observational and experimental field studies in a range of systems have failed to find a straightforward relationship between temperature and potential enzyme activity (Bonnett et al. 2006; Bell et al. 2010; Bell and Henry 2011; Henry 2012). In our study, a similar result is perhaps not surprising given the modest amount of experimental warming (+1 °C). However, temperatures increases of similar magnitude translate to increased rates of soil respiration and nitrogen turnover in a range of systems (Rustad et al. 2001). Moreover, in our experimental system, both of these fluxes have been observed to increase dramatically (50–100 % increase) under the same modest warming regime (Dorrepaal et al. 2009; Weedon et al. 2012). The corresponding lack of detectable changes in soil enzyme pools imply that these effects are not caused by changes in the pool sizes of the enzymes involved in the transformations. One explanation for this discrepancy is that enzyme activity assays measure potential activity under non-limiting levels of substrate (Allison et al. 2007). It is therefore possible that warming effects on C and N fluxes could be related to changes in the supply of substrate, as this can lead to large differences in in situ turnover rates of C and N without differences in the measured potential activity of the enzymes involved in the fluxes (Weintraub and Schimel 2005). Another, nonexclusive, explanation is that other enzyme pathways, not measured in this study, are more closely linked to in situ flux rates than the glycosidases and amino-peptidases we measured in this study. For nitrogen cycling, chitinases and endo-protease enzymes may also be important (Caldwell 2005). In the carbon cycle, oxidative enzymes may be the key control on flux rates (Freeman et al. 2001). More generally, the extent to which we can link enzyme measurements to variation in environmental conditions and ecosystem processes of interest remains a key challenge in soil enzymology (Allison et al. 2007; Weedon et al. 2011; Henry 2012).

Seasonality of enzyme pool sizes and temperature sensitivities

Despite the lack of effects of climate treatments, the strong seasonality in the relative sizes of enzyme pools suggests that they are subject to some degree of regulation. Enzyme pools at a given moment in time are the net result of processes of production, degradation and immobilization. Thus, we cannot definitively assert that this regulation is actively exerted by the microbial community, or even if we could, whether this regulation occurs at the level of changes in the composition of the enzyme-producing community (Allison and Martiny 2008), or through shifts in the production of enzymes within an otherwise stable community. However, the changes in temperature sensitivity can be interpreted as a shift in the isoenzyme composition of the enzyme pools (Wallenstein et al. 2011), where isoenzymes are groups of enzymes with the same catalytic function but different structure. If this is the case it could be considered as evidence in favour of microbe-level control of enzyme pools. For the purposes of the following discussion, we assume that the patterns we have observed are adaptive—and can be interpreted in terms of the nutritional status of the soil microbial community.

Control of extracellular enzyme production by soil microbes can be understood in terms of ecological-economic constraints (Allison et al. 2011). Microbes should allocate resources to specific enzymes that allow them to obtain resources which are limiting growth and reproduction. At the same time, substrate-induction of enzyme production could also be important, as there is no energetic benefit in producing enzymes for which the corresponding substrate is not present in the surrounding environment (Allison and Vitousek 2005; German et al. 2011). In this framework, if we assume peptidase production is driven primarily by N demand, the higher potential activities of these enzymes in June relative to August may reflect a greater relative demand for nitrogen earlier in the season. This is supported by the relatively lower concentrations of inorganic N in June (Weedon et al. 2012). This interpretation may also help explain the corresponding increase in beta-glucosidase activity across the growing season. The microbial biomass could be shifting from demand for nitrogen in June to demand for carbon/energy in August, with corresponding shifts in the relative production of peptidases and glucosidases (the implications of this temporal divergence of C and N dynamics for modeling are explored further below). However, the data for the other C cycle enzymes suggest a more complex picture. In terms of pool sizes, alpha-glucosidase shows a magnitude of relative decline comparable to that seen for the four peptidases. A possible explanation is that the production of this enzyme is controlled more by substrate supply than demand, as was demonstrated in experiments demonstrating a positive relationship between starch concentration and alpha-glucosidase activity (German et al. 2011). More generally, the contrasting patterns between beta-glucosidase and alpha-glucosidase potential activities may indicate a shift in allocation from enzymes degrading more labile C sources (starch) early in the season towards cellulose metabolism later in the season when labile C sources are scarcer.

It is important to bear in mind that the foregoing interpretation of seasonal dynamics is based on only two sampling occasions, a consequence of the need to limit destructive sampling in a long-term, multi-use experiment. The seasonal patterns we tentatively identify would need to be validated by more intensive temporal sampling, preferably in a wider variety of systems. Nevertheless, our interpretation broadly agrees with the findings of Weintraub and Schimel (2005), who concluded that an early season peak in protease activity in tundra soils was driven by increased microbial demand for N and speculated that this microbial N limitation was partly driven by a transient increase in the supply of labile C by plants. In general, seasonally frozen soils, such as those in (sub-) arctic and alpine environments are characterized by a strong seasonality in the availability, mobilization, and uptake of carbon and nitrogen (Falge et al. 2002). This is a consequence of rapid turnover of microbial biomass during spring-thaw (Schmidt et al. 2007), and the temporal dynamics of plant nutrient uptake and rhizodeposition during the relatively short plant growing season (Jaeger et al. 1999). Thus, the strong seasonality in substrate supply and enzyme activities seen in this and other studies (Bonnett et al. 2006; Wallenstein et al. 2009; Bell et al. 2010) seems to be a general characteristic of cold and temperate climates. More studies in Mediterranean, (semi-) arid and tropical environments would help to determine if similar patterns exist in warmer biomes.

While we have focussed on the role of substrate supply, other environmental factors could potentially produce the observed seasonal patterns in enzyme measurements. Firstly, variations in soil moisture have been related to enzyme activity measurements in other systems (Henry 2012). At our site, soil moisture showed only small declines (from 88.0 to 86.9 % by weight) over the entire study period, a fluctuation that is probably not sufficient to explain the patterns we observed. Secondly, soil temperature increased over the course of the growing season (Fig. 1) and could be invoked to explain the seasonal changes in enzyme potential activity. However, we propose that it is more likely that temperature acts on enzyme pools indirectly, i.e. by driving the seasonality of carbon and nutrient availability as discussed above. Experimental incubations of peat soils showed that enzyme pools were much more sensitive to substrate supply than to even a 5 °C increase in incubation temperature (Weedon et al. 2013), and, as previously noted, experimental field studies in a range of systems (including the present study) have failed to find a straightforward relationship between temperature and potential enzyme activity (Bell et al. 2010; Henry 2012).

As well as providing evidence for adaptive control over enzyme production (see above), our observation of contrasting shifts in the temperature sensitivity of Vmax over the season may complicate our understanding of warming effects on soil stoichiometric balance. Previous studies have found a higher temperature sensitivity of glycosidases compared to peptidases, and suggested that this discrepancy could have consequences for the balance between C and N cycles under global warming (Koch et al. 2007; Wallenstein et al. 2009; Wallenstein et al. 2011). Our results suggest that the relative temperature sensitivity of C-/N-related processes may shift across the growing season, and thus extrapolating to global warming effects will require more detailed knowledge about the timing of important “hot moments” in soil C and N cycling.

Implications of contrasting temporal patterns in C and N enzymes

As with all studies based on a single environment, observations in a broader range of ecosystems are needed to test the generality of the patterns we have observed. This may be particularly relevant in this case given the somewhat atypical nature of peatland soils: low pH and nutrient availability, high organic matter content and lack of a mineral matrix (Vitt 2006). However, if the seasonal patterns in enzyme pool sizes and composition we have described are found to be more general, it follows that there are some important implications for the interpretation of soil enzyme measurements, and our understanding of the relationship between those measurements and the stoichiometry of soil nutrient and energy cycles.

Recent models for soil carbon and nitrogen dynamics that include microbial biomass and enzyme production assume that enzyme production is a function of microbial biomass, and usually do not distinguish between enzymes produced to acquire carbon/energy, and those produced to acquire nutrients (e.g. nitrogen and phosphorus) (Schimel and Weintraub 2003; Moorhead and Sinsabaugh 2006; Allison et al. 2010). This assumption is supported by recent global syntheses of enzyme activity measurements in soils and sediments which show a convergent 1:1:1 stoichiometry of allocation to enzymes acquiring C, N and P (Sinsabaugh et al. 2008; Sinsabaugh et al. 2009). However, the extent to which this relationship applies to the small spatial and temporal scales relevant to local nutrient cycling remains an open question. Patterns of enzyme stoichiometry that emerge from such global, intersite comparisons over large gradients may be dominated by top-level (e.g. climate, productivity) constraints on enzyme production, and mask significant divergence at the local scale—i.e. within sites, and across seasons—such as that observed in the present study.

Although parsimonious and tractable model frameworks are important, if the seasonal divergence we have observed is more widespread, strategies for incorporating differential potential activity and temperature sensitivity of enzyme functional groups should be developed and their influence on model behaviour assessed. The recently published EEZY model of Moorhead et al. (2012), and DEMENT model of Allison (2012) are examples of approaches that model separate enzyme pools for C and N acquisition (roughly corresponding to the glycosidases and peptidases measured in the present study). These models are able to successfully re-create observed patterns of enzyme dynamics and decomposition rates. Extending this approach to include empirically calibrated temporal variation in environmental drivers and substrate supply could be a useful means for exploring drivers of the seasonal dynamics of soil processes and, in turn, identify critical questions for further field investigations. Such a process of constant feedback between empirical observation and modelling efforts (as exemplified by this special issue) holds the potential to yield great progress in illuminating the role of soil enzymes in soil functioning, and their response to global change.

Conclusions

We found no evidence for effects of increased soil temperatures on the potential activities or temperature sensitivities of glycosidases and peptidases in a sub-arctic peatland, despite previously observed large effects of higher temperatures on respiration and N cycling. On the other hand, we saw a strong seasonal pattern in both enzymatic parameters with opposite seasonal trends for the two functional groups. Our results suggest that warming effects on soil processes in this globally important carbon sink are more likely to act indirectly via patterns in substrate supply, rather than directly via changes to the production of soil enzymes. Furthermore, the observed divergence of glycosidase and peptidase enzyme activities and temperature sensitivities on a seasonal timescale suggest that modelling approaches that treat these functional groups separately and partly independently should be further explored. Deeper mechanistic understanding of the complex interplay of C and N cycles in soil ecosystems will allow more robust predictions of the response of these systems to climate change.