Increasing soil temperature has the potential to alter the activity of the extracellular enzymes that mobilize nitrogen (N) from soil organic matter (SOM) and ultimately the availability of N for primary production. Proteolytic enzymes depolymerize N from proteinaceous components of SOM into amino acids, and their activity is a principal driver of the within-system cycle of soil N. The objectives of this study were to investigate whether the soils of temperate forest tree species differ in the temperature sensitivity of proteolytic enzyme activity over the growing season and the role of substrate limitation in regulating temperature sensitivity. Across species and sampling dates, proteolytic enzyme activity had relatively low sensitivity to temperature with a mean activation energy (Ea) of 33.5 kJ mol−1. Ea declined in white ash, American beech, and eastern hemlock soils across the growing season as soils warmed. By contrast, Eain sugar maple soil increased across the growing season. We used these data to develop a species-specific empirical model of proteolytic enzyme activity for the 2009 calendar year and studied the interactive effects of soil temperature (ambient or +5°C) and substrate limitation (ambient or elevated protein) on enzyme activity. Declines in substrate limitation had a larger single-factor effect on proteolytic enzyme activity than temperature, particularly in the spring. There was, however, a large synergistic effect of increasing temperature and substrate supply on proteolytic enzyme activity. Our results suggest limited increases in N availability with climate warming unless there is a parallel increase in the availability of protein substrates.
 Extracellular enzyme activity is thought to be highly temperature sensitive [Allison et al., 2010]; as temperature rises, the rate of substrate diffusion increases and sufficient energy becomes available for the catabolism of substrates by enzymes (i.e., overcoming a substrate's activation energy). The environmental conditions under which enzymes are produced, also affects their reactivity. Enzymes produced in warm soils tend to be less temperature sensitive than enzymes produced in cold soils as a result of an increase in enzyme rigidity with temperature, which slows the rate of enzyme binding to substrates and the release of products [Fenner et al., 2005; Koch et al., 2007]. The tradeoff, however, is that enzymes operating at warmer temperatures retain structural integrity and operational capacity at higher temperatures [Somero, 1978]. In cold soils, by contrast, enzymes tend to have higher temperature sensitivities because their flexibility improves the conformation between enzyme and substrate, thus resulting in greater potential activity at low temperatures [Siddiqui and Cavicchioli, 2006; Bradford et al., 2010].
 Temperate forest tree species of the northeastern U.S. differ in many of the factors that control the activity of extracellular enzymes [Lovett et al., 2004]. The species in these forests differ in litter chemistry (C:N, lignin:N) [Finzi et al., 1998] and mycorrhizal fungal association (i.e., ecto- versus arbuscular mycorrhizae), which leads to significant interspecific variations in rates of microbial activity and N cycling [Finzi et al., 1998, Brzostek and Finzi, 2011]. These species also differ in the concentrations and chemistries of polyphenolic compounds in litter. In particular, variation between these species in tannin chemistry (i.e., the ratio of condensed to hydrolysable tannins) can lead to differences in the chemical protection of protein in SOM [Kraus et al., 2003a, 2003b]. In northeastern U.S. forests, tree species that form associations with ectomycorrhizal (ECM) fungi have recalcitrant leaf litter that is dominated by condensed tannins; whereas the leaf litter of arbuscular mycorrhizal (AM) associated trees is relatively labile and dominated by hydrolysable tannins [Talbot and Finzi, 2008]. Condensed tannins chemically protect protein to a greater extent than hydrolysable tannins and may limit the availability of protein substrates for enzymatic attack [Talbot and Finzi, 2008].
 To understand the aggregate effects of soil temperature and substrate supply on proteolytic rates, we quantified seasonal changes in the temperature sensitivity of proteolytic enzyme activity with and without the addition of protein substrate in soils influenced by four common tree species in the eastern deciduous forests of North America. We then used this information to develop a model of proteolytic enzyme activity across the calendar year taking into account seasonal variations in the temperature sensitivity and substrate limitation of proteolytic enzyme activity. We tested three hypotheses: (1) proteolytic enzyme activity is more temperature sensitive in cold soils than in warm soils, (2) proteolytic enzyme activity is more temperature sensitive in soils with labile leaf litter inputs (i.e., low lignin and condensed tannin concentrations, high N) than soils with recalcitrant leaf litter inputs due to greater supply of substrate for enzymatic attack, and (3) proteolytic enzyme activity is more strongly limited by substrate availability than temperature.
2. Materials and Methods
2.1. Site and Plot Description
 This research was conducted at two sites, one located at the Harvard Forest (HF) in Petersham, Mass. (42.5°N, 72.18°W), and the other at Pisgah State Forest (PSF) in Chesterfield, N. H. (42.87°N, 72.45°W). The sites have similar land-use history and stand age [Foster, 1988, 1992]. Soils at both sites are inceptisols classified as Typic Dystrochrepts derived from glacial till overlying granite-schist-gneiss bedrock (USDA National Resource Conservation Service,http://websoilsurvey.nrcs.usda.gov/).
 Study plots dominated by one of four target tree species were established at each site. Stands of sugar maple (Acer saccharum) and American beech (Fagus grandifolia) were located in the PSF. Stands of eastern hemlock (Tsuga canadensis) and white ash (Fraxinus americana) were located in the HF. These four species differed in mycorrhizal association, with white ash and sugar maple supporting arbuscular mycorrhizal (AM) fungi and hemlock and beech supporting ectomycorrhizal (ECM) fungi. At each site we located six replicate, 8 m radius monodominant plots in which the target tree species constituted 100% of standing basal area in the inner 5 m core and >80% in the entire 8 m radius plot [Lovett et al., 2004].
 Soil samples were collected in April, June, and August of 2009 from each of the six replicate plots for each species (n = 24). We chose the April, June, and August time points because they span the majority of the growing season, including leaf out, peak photosynthesis and leaf area index, and the start of the seasonal decline in C uptake at the HF, respectively [Urbanski et al., 2007]. At each sampling point, we collected one sample of the top 15 cm of mineral soil using a 5 cm diameter soil bulk-density sampler from each plot. We focused on the mineral soil because only hemlock and beech had appreciably organic horizon accumulations. The samples were processed within 12 h of collection in the laboratory. Fine roots were removed from each sample, and the sample was then sieved through 2 mm mesh.
2.2. Temperature Sensitivity of Proteolytic Enzyme Activity
 To test hypothesis 1, that proteolytic enzyme activity is more temperature sensitive in cold soils than in warm soils, proteolytic rates in the top 15 cm of soil collected in April, June, and August were assayed in the lab at ambient substrate levels and at six temperatures: 4°C, 10°C, 17.5°C, 23°C, 30°C, and 35°C. Four replicate 2–3 g subsamples of soil from each sample were placed in their proper incubation temperature and allowed to equilibrate for 12 h. Based on preliminary studies, the 12 h equilibration period was required for all samples to reach the correct assay temperature.
 After equilibration, we assayed proteolytic enzyme activity at the target incubation temperature using a method modified from Watanabe and Hayano  and Lipson et al.  and described in Berthrong and Finzi . This assay measured the gross flux of amino acids in the absence of plant and microbial uptake. In brief, soils were incubated in a sodium acetate buffer with a small volume of toluene (400 μl) to inhibit microbial uptake by increasing the permeability of microbial cell walls. After the reagent was added, 2 of the 4 replicate samples were incubated for 15 min and then extracted in 3 ml of trichloroacetic acid solution. The 2 remaining samples were incubated for 4 h and then extracted in 3 ml of trichloroacetic acid solution. Previous work has shown that proteolytic enzyme activity is linear over this time period in temperate forest soils [Berthrong and Finzi, 2006; Rothstein, 2009]. The purpose of the samples extracted after 15 min (i.e., the initial samples) was to account for free pools of amino acids in the soil in addition to any flush of amino acids from permeable microbial cell membranes resulting from the addition of toluene to each sample. Proteolytic rates for each experimental treatment were calculated as the difference between amino acid concentration in the paired incubated and initial samples. The concentration of amino acids for all samples was quantified using the o-phthaldialdehyde andβ-mercaptoethanol (OPAME) method [Jones et al., 2002]. Concentrations of amino acid N were determined by comparing the fluorescence of the samples relative to a standard curve composed of glycine.
 Recently, Reiskind et al.  renewed an older debate [see Frankenberger and Johanson, 1986] on the artifacts of toluene use in enzyme assays. Toluene is commonly used because it increases the permeability of microbial cell membranes and inhibits the active uptake of enzyme products by microbes [Frankenberger and Johanson, 1986]. In the temperate forest soils studied here toluene addition is essential for assaying activity; in the absence of its addition, we have found only amino acid immobilization (data not shown). While there are a number of potential artifacts with the use of toluene (e.g., diffusion of intracellular products and enzymes into solution and the potential for the inhibition of enzyme activity) [Deng and Tabatabai, 1994; Nannipieri et al., 1996; Hofmockel et al., 2007], the data derived from these studies are very useful. For example, Lipson et al.  parameterized a model from proteolytic rates assayed with toluene in one year and found that the model accurately predicted amino acid availability in the subsequent year in an alpine ecosystem. Moreover, the procedure for assaying proteolytic activity in this study minimized the potential for experimental artifacts. The initial samples were dosed with toluene to control for the release of intracellular amino acids, which is likely to occur immediately after addition given the diffusion gradient set by the concentration of amino acids within microbial cells relative to that in soils. The four hour incubation also minimized the contribution of intracellular processes to activity while still ensuring that activity was linear [Berthrong and Finzi, 2006]. Finally the same technique was used for all samples, so to the extent that there are artifacts, the bias is similar among all samples. Thus, while the use of toluene as a bacteriostatic agent may be imperfect, it is a very useful tool for measuring proteolytic enzyme activity at ambient substrate conditions given the inherent difficulties in measuring protein substrate depletion in soil [Roberts and Jones, 2008] and the reliance of other methods on the addition of substrate (i.e., the use of fluorometrically linked or isotopically labeled substrates).
 We modeled proteolytic activity as a function of temperature (4°C to 35°C) using the Arrhenius equation [Arrhenius, 1889]:
where R is the universal gas constant, T is the temperature in Kelvin, Ais the pre-exponential factor, andEa is the activation energy of the substrate. We fit this equation for a given soil sample from each plot for every month, separately (n = 72). The Arrhenius equation provided a good description of the response of proteolytic enzyme activity to changes in temperature for each sampling date (overall average R2 = 0.84) and allowed us to measure seasonal changes in the apparent activation energy, Ea, at the ambient protein substrate supply. On a number of sample dates in April and June, the higher temperature treatments (30°C and 35°C) had lower rates of proteolytic enzyme activity than the next closest (cooler) temperature. When this occurred, the temperature treatment was removed from this analysis.
 We used two-way ANOVA to test for the effects of species and sample date on theEaof proteolytic enzyme activity. To examine seasonal effects for a given species, we used one-way ANOVA to test for the effect of sample date on theEavalues. The data were analyzed with SAS, version 9 (SAS Institute, Cary, N. C.) using the GLM procedure. Data were assessed for normality and homogeneity of variance and log-transformed to meet model assumptions when needed. Tukey's Studentized Range Test was used for all post hoc comparisons of mean differences among species or sample date.
2.3. Modeling Proteolytic Enzyme Activity
 We developed a model of proteolytic enzyme activity in the soil of each tree species. This model uses an approach similar to that of Davidson et al. , where Arrhenius kinetics are used to model the effects of temperature and Michaelis–Menten kinetics are used to model the effects of substrate availability on enzyme activity. We developed the model to evaluate the interactive effects of substrate availability and temperature on seasonal variation in proteolytic enzyme activity, and to assess how warming, here a 5°C increase in mean annual soil temperature, affects proteolytic enzyme activity.
 The baseline model was developed under ambient temperature and substrate conditions using the data collected in April, June and August of 2009. To extend the baseline model to the entire calendar year we used the 2009 daily record of soil temperature measured in hemlock and hardwood stands at the HF (A. Ellison, Harvard Forest Data Archive: HF108). Using the Arrhenius equation (equation (1)), we modeled proteolytic enzyme activity at a daily time step by linearly interpolating A and Ea between the April, June and August sample dates. Outside of the measurement period, daily Ea values were estimated from the empirical relationship between Ea and soil temperature measured during the growing season (Figure 1c), and the pre-exponential factor,A, estimated from the positive relationship between Ea and A (R2 = 0.99).
 To model the effect of a 5°C rise in soil temperature, we used the equation:
We generated Te by increasing the daily soil temperature by 5°K for each calendar day. The A and Ea parameters in equation (2) were assumed to be the same as those in the baseline model.
 It was not possible to measure the temperature sensitivity of proteolytic enzyme activity under elevated substrate conditions across the entire range of temperatures described above; owing to high replication, this would have generated a prohibitively large number of samples. Rather, we used previously published data to establish a model of proteolytic enzyme activity at ambient and elevated substrate.
 The model of the effect of elevated substrate on proteolytic activity was based upon the Michaelis–Menten equation:
where y is proteolytic enzyme activity at ambient substrate, [protein] is the concentration of available substrate, Vmax is the maximum velocity of the reaction at saturating substrate conditions, and Km is the Michaelis–Menten constant describing the substrate affinity of the enzyme.
 To parameterize the model, we first used the previously published data to estimate seasonal variation in Km using equation (3). The values for [protein] are from Rothstein , who measured the monthly concentration of soluble peptides in soils dominated by ECM and AM tree species from April to September 2005 (Figure 1). We assumed that the values of y and Vmax were equivalent to proteolytic enzyme activity assayed at ambient and saturating substrate conditions, respectively, using the data published in Brzostek and Finzi  for April, June and August of 2008, and in Berthrong and Finzi  for October and November 2002. In both of these studies, activity was assayed at ambient conditions as described above, and under saturating substrate conditions by adding casein to the buffer solution (0.6% casein) [Lipson et al., 1999]. With these data, we solved for Km in equation (3) and thus estimated the seasonal changes in Km between April and October (Figure 1). Importantly, the data from Brzostek and Finzi  were generated from the same plots studied here. The data from Rothstein  and Berthrong and Finzi  are from similar cold temperate forests in Michigan and Connecticut, respectively, all of whom sampled soils from either ECM or AM dominated stands.
 To model proteolytic enzyme activity at elevated or nonlimiting substrate conditions (i.e., Vmax), under ambient temperatures, we used a modified version of equation (3):
where y is the daily estimate of proteolytic enzyme activity from equation (1), [protein] is from Rothstein , and Km is the estimate generated above. As with our estimate of Ea in equation (1), we linearly interpolated values for Km and [protein] between the measurement dates—Day of Year (DOY) 100–312—and made conservative assumptions outside of the measurement period.
Km was assumed to equal the concentration of available substrate (i.e., [protein]) when soil temperatures were below 1°C, corresponding to DOY 1–84 and 357–365. In the early spring (DOY = 85–100), we assumed that Km increased linearly through time to the measured value in April. In the fall and early winter (DOY = 312–356) we assumed that Km increased linearly from the measured value in November to the size of the available [protein] pool as the soil approached 1°C in the winter. We assumed that Km equaled [protein] in winter soils to provide a conservative annual estimate of the importance of substrate limitation. We made this assumption because we had little information on how Km responds in winter soils and because enzymes typically have Km's that mirror substrate availability [Somero, 1978]. Similarly, we assumed that from DOY 1–143, [protein] equaled the measured value in the April from Rothstein . From DOY 269–365, we assumed that [protein] increased linearly through time to the measured value in April.
 To model the response of proteolytic enzymes to elevated substrate, Vmax, under elevated temperatures, we used the same approach and values for Km and [protein] as described above for ambient temperatures. The lone difference is that we used the daily estimate of proteolytic enzyme activity at Te from equation (2) for y in equation (4).
 The model was used to test hypothesis 2, that proteolytic enzyme activity is more strongly limited by substrate availability than temperature, by comparing seasonal and annual rates of amino acid production under factorial manipulations of temperature and substrate supply. The four modeling scenarios were as follows: (1) ambient temperature and substrate availability, (2) elevated temperature (+5°C across the entire year) and ambient substrate availability, (3) ambient temperature and elevated substrate availability, and (4) elevated temperature and elevated substrate availability. To calculate the annual rate of gross amino acid production, we assumed no diurnal variations in proteolytic enzyme activity and scaled the 4 h rates measured in the lab to daily rates of proteolysis. We then summed daily rates of proteolytic enzyme activity across the entire year.
 Like all models, this activity required that we make a number of assumptions because of the need to extrapolate from our point-in-time estimates of processes to the entire year that add to the uncertainty in the model. For all the model simulations, we have strived to be transparent and conservative in these assumptions. Most of the assumptions in the model were made during the winter and highlight the need for the scientific community to expand investigations of SOM decomposition into the winter months. To parameterize the substrate model across a broad span of the growing season, we needed to use data from different geographical locations and years.
 Finally, the previously published data on the activity of proteolytic enzyme under saturating conditions was assayed using casein, an animal protein, which is commonly used in soil assays to stimulate an unbound, unprotected protein [e.g., Lipson et al., 2001; Weintraub and Schimel, 2005; Hofmockel et al., 2010]. There is the potential that casein may be more easily degraded than the in situ soil protein pool because it lacks tertiary structure. However, the most common plant protein in leaves, Rubisco, is actively degraded by plants during senescence and the peptides and proteins that reach the soil lack the complex structure of the parent protein [Feller et al., 2008].
3.1. Temperature Sensitivity of Proteolytic Enzyme Activity
 The maximum rate of proteolytic enzyme activity shifted from 23°C in April to 30°C in June to 35°C in August in ash and hemlock soils (Figures 2a and 2d). Beech soils showed a similar trend, except that activity reached its maximum at 30°C in April due to a marginal increase in activity from 23 to 30°C (Figure 2b). In sugar maple soils, proteolytic enzyme activity reached its maximum at 35°C in April and August (Figure 2c). In June, the maximum rate of proteolytic activity occurred at 30°C (Figure 2c).
 The apparent activation energy of the protein substrate ranged from 18 to 60 kJ mol−1 (Table 1). On average, protein pools in sugar maple soils had the highest Ea and hemlock the lowest (p < 0.05) (Table 1), though the species rankings varied subtly among the three sampling dates. Seasonally, there was a significant interaction between species and sampling date on Ea (p < 0.005). In ash, hemlock, and beech soils, Ea declined across the growing season, though significantly (p < 0.05) only in ash and beech. The opposite was observed in sugar maple soils; Ea's in August were significantly higher than those in June and April (p < 0.05) (Table 1).
Table 1. Mean (±1 SE) Ea Values for Each Sample Date and Across the Growing Seasona
Different uppercase superscript letters indicate significant differences (p < 0.05) within a given species between months. Different lowercase letters indicate significant differences (p < 0.05) between species in the growing season average.
3.2. Modeled Proteolytic Enzyme Activity
 Proteolytic rates were low in the winter, increased with increasing soil temperature in the spring and early summer, and decreased as soil temperature declined in the fall and winter. The peak rate of proteolytic enzyme activity in sugar maple soils was ∼30% greater than in the ash, hemlock, and beech stands (Figures 3a–3d). In sugar maple, ash, and beech soils, the peak occurred at midsummer (DOY 235), when soil temperatures reached their maximum (Figures 3a, 3c, and 3d). In hemlock soils the peak in enzyme activity occurred earlier, on DOY 211 (Figure 3b). Integrating over the entire year, hemlock soils had the highest rates of proteolytic activity, which were largely the result of higher enzyme activity at low winter temperatures (Table 2 and Figure 3a). Despite having the highest peak rate, sugar maple had the lowest annual rate as a result of low activity over the winter months (Table 2 and Figure 3c).
Table 2. Modeled Annual Rates and the Proportional Stimulation of Proteolytic Enzyme Activity Under Each Scenario for the Soil of Each Speciesa
Annual Rates (mg AA-N g dry soil−1 yr−1)
Elevated T × S
Elevated S × T
Elevated temperature (T), elevated substrate (S), and the elevated T × S proportional stimulations are the ratio of the annual rate of proteolytic enzyme activity for each elevated scenario divided by the ambient annual rate of proteolytic enzyme activity.
 The yearly pattern of proteolytic enzyme activity at elevated soil temperature (+5°C) mirrored that at ambient temperature, though there were interspecific differences in the timing of the response to temperature. In hemlock and beech soils, the greatest increase in activity occurred during the late winter into early spring (Figures 3a–3b and Figure 4). There was little effect of warming on summer rates of proteolytic activity because warming accelerated the decline in the temperature sensitivity of enzyme activity across the growing season (Figure 4). By contrast, warming increased proteolytic activity over the growing season in sugar maple and ash soil (Figures 3c–3d), because of their high temperature sensitivity throughout the growing season (Table 1). Across the year elevating temperature by 5°C stimulated proteolytic enzyme activity 19%–36% (Table 2), with the largest stimulation in ash and sugar maple soil and the smallest in hemlock soil.
 The addition of protein substrate increased annual proteolytic enzyme activity 81%–193% above that observed at ambient substrate supply and temperature (Table 2). In contrast to the effect of temperature, there was little difference among the species in the timing of the response to substrate addition. In all of the soils, the substrate-induced stimulation of proteolytic activity peaked early in the growing season when the soils were still cold (Figures 3e–3h). As soils warmed, the stimulation in proteolytic activity declined substantially. However, there was a strong secondary peak in substrate limitation due to a sharp decline in substrate availability on DOY 208 in hemlock and beech soils and DOY 269 in maple and ash soils (Figures 1a–1b and Figures 3e–3h).
 There was a large, synergistic effect of increasing substrate supply and soil temperature on proteolytic enzyme activity; rates increased 136%–297% above the baseline model (Table 2 and Figures 3e–3h). This synergistic effect was greatest during the spring peak in activity for all the soils (Figures 3e–3h).
 The effect of elevated substrate supply on proteolytic activity at both temperatures was largest in the sugar maple and ash soils and smallest in hemlock and beech soils (Figures 3e–3h). This pattern was driven primarily by sugar maple and ash soils having higher Km's on average than beech and hemlock soils (Figures 1a–1b).
 There were large seasonal differences in the temperature sensitivity of proteolytic enzyme activity (Figure 2). Across species and sampling dates, the Ea of proteolytic enzyme activity varied from 18 to 60 kJ mol−1 (Table 1 and Figure 2). This measure of temperature sensitivity is, on average (mean Ea = 33.5 kJ mol−1), less than half the value reported for the temperature sensitivity of C mineralization [e.g., Fierer et al., 2006; Craine et al., 2010]. In the model, the low activation energies led to a modest increase in proteolytic enzyme activity as soils warmed from the winter through the summer, or in response to experimental warming (Figures 3a–3d). Proteolytic enzyme activity was considerably more responsive to increases in protein substrate availability than soil temperature alone (Table 2 and Figure 3). The response of proteolytic enzyme activity to elevated temperature and substrate supply was, however, substantially greater than that of elevated substrate alone indicating interactive effects of temperature and substrate supply on enzyme activity (Table 2). The results of this study suggest that soil warming as a result of climate change will have only a modest effect on the depolymerization of N into amino acids unless there is a concomitant increase in the mobilization of protein substrate from SOM.
 The seasonal decline in the apparent activation energy of proteolytic activity in ash, hemlock, and beech soils supports hypothesis (1), that proteolytic enzyme activity is more temperature sensitive in cold soils than in warm soils. The seasonal reduction in the maximum proteolytic activity at optimal substrate conditions (i.e., Vmax) indicates that the quantity of active enzymes decreases across the growing season (Figure 3) [Brzostek and Finzi, 2011; Wallenstein et al., 2009]. Also, the seasonal decline in enzyme activity was synchronous with the springtime peak and subsequent decline in soluble peptide concentrations in the soil (Figure 1) [Rothstein, 2009]. This suggests that substrate limitation decreases microbial investment in the production of proteolytic enzymes, and hence the response of proteolytic activity to temperature.
 An interaction between leaf litter chemistry and the dynamics of protein protection by SOM may explain the greater activation energy of proteolytic enzyme activity in AM compared to ECM soil (Table 1). In particular, AM litter inputs are dominated by hydrolysable tannin whereas ECM litter inputs are dominated by condensed tannin [Talbot and Finzi, 2008]. Although both types of tannins have the ability to bind proteins, condensed tannins protect protein from decomposition to a much greater degree than hydrolysable tannins [Kraus et al., 2003b; Bowman et al., 2004], and condensed tannins also bind soil enzymes, reducing activity [Joanisse et al., 2007]. Thus in AM soils, the greater apparent temperature sensitivity of proteolytic enzyme activity may be fueled by a greater availability of weakly bound SOM protein (Table 1). This line of argumentation could also help explain the increase in proteolytic activity across the growing season in sugar maple soil, the only species to display this pattern (Table 1).
 Consistent with hypothesis (3) we found that substrate availability strongly limited proteolytic enzyme activity (Table 2). In all of the soils, substrate limitation peaked in the spring and then declined as the growing season progressed (Figures 3e–3h). Again, changes in the enzyme pool size appear to control this pattern. Wallenstein et al.  found a strikingly similar pattern in Arctic soils and proposed that slow rates of proteolytic enzyme degradation in cold winter soils results in a large pool of active enzymes in the spring. Further, the growing season decline in proteolytic activity at optimal substrate conditions suggests that the microbial synthesis of proteolytic enzymes becomes increasingly limited by the availability of nutrients or is down regulated as a result of a decline in microbial N limitation [Geisseler and Horwath, 2008; Allison et al., 2009; Wallenstein et al., 2009].
 There were three primary sources of uncertainty in our model of proteolytic enzyme activity over the calendar year. Importantly, these sources of uncertainty provide a framework for future research on proteolytic enzyme kinetics. First, the data used to parameterize the model was collected only during the growing season and may not be reflective of proteolytic enzyme kinetics in winter soils. In particular, there can be substantial rates of microbial activity in the wintertime, which has very high apparent temperature sensitivity [e.g., Mikan et al., 2002; Koch et al., 2007; Wallenstein et al., 2009]. Second, there is uncertainty in our estimates of the model inputs (e.g., Ea, Km, [protein]). Finally, our model does not consider the effects of soil moisture on the solubility of enzymes and substrates, though the years in which we sampled soils, the Harvard Forest received above average precipitation (1565 mm in 2008; 1277 mm in 2009; Boose E, Harvard Forest Data Archive: HF001). Future efforts to model proteolytic enzyme activity, particularly in xeric to mesic ecosystems, should explicitly consider the effects of soil moisture.
 To address whether the modeled estimates of proteolytic rates in this study underestimated over-winter proteolytic enzyme activity, we conducted an additional simulation experiment in which the temperature sensitivity of proteolytic enzyme activity was doubled in winter months (i.e.,Ea*2). Doubling the winter temperature sensitivity of proteolytic enzyme activity increased amino acid production 4%–7% more than the 5°C warming simulation without modification of the winter Ea value (Table 3). The reason for the modest contribution is the low baseline activity of these enzymes at temperatures <5°C (Figure 2), suggesting that the greater temperature sensitivity of wintertime enzyme activity is of less importance to the annual production of amino acids than changes in activity and substrate supply in the spring and early summer in productive, temperate forests.
Table 3. Modeled Annual Rate of Proteolytic Enzyme Activity with a Doubling of the Winter Ea and Elevated Temperature Conditions for the Soils of Each Tree Speciesa
The annual rates are in units of mg AA-N g dry soil−1 yr−1. The proportional stimulation is the ratio of the annual rate of proteolytic enzyme activity under elevated T conditions with a doubling of winter Ea to the elevated T model without an Ea manipulation.
 We also tested the sensitivity of the model to the measured error in the model inputs (i.e., Ea, Km, and [protein]) (Table 4). For simplicity, we focused on the ambient temperature model at ambient and elevated substrate. The model was more sensitive to variation in the Km, and [protein] than Ea (Table 4). In particular, reductions in [protein] drove a logical increase in substrate limitation for all the species. Even with the added uncertainty, the range of annual estimates at elevated substrate was still substantially higher than those at ambient substrate conditions (Table 4).
Table 4. Sensitivity of Modeled Annual Rates of Proteolytic Enzyme Activity to ±1SE Variation in Ea Values in the Baseline Model and to ±1SE Variation in Km and [protein] in the Elevated Substrate Modela
The annual rates are in units of mg AA-N g dry soil−1 yr−1.
 The amino acids produced by proteolytic enzymes contribute to temperate forest productivity and serve as substrates for mineralization and nitrification [Finzi and Berthrong, 2005; Gallet-Budynek et al., 2009]. As such, proteolytic enzyme activity is a key component of the N cycle in temperate forests. The overall low temperature sensitivity and the strong substrate limitation of proteolytic enzyme activity suggest that the positive feedback of warmer temperatures on N cycling may not be as large as formulated in the current generation of biogeochemical models [e.g., Sokolov et al. 2008; Thornton et al. 2009; Zaehle et al. 2010]. Further, seasonal changes in the concentration of enzymes and substrates in soils appear to be the dominant control on the response of proteolytic enzymes to temperature. Thus, to improve our understanding of how the N cycle will respond to global change, future studies should focus on the processes controlling the microbial synthesis of enzymes and the mobilization of protein from SOM.
 We would like to thank Colin Averill, Joy Cookingham, Catherine Achorn, Verity Salmon, Poliana Lemos, and Branden Rider for laboratory and field assistance. We thank Andy Reinmann, John Drake, and the anonymous reviewers for their comments on an earlier version of this manuscript. In addition, we thank the Harvard Forest and the New Hampshire Department of Resources and Economic Development for granting us access to perform field research. This work was supported by a grant from the National Science Foundation (DEB-0743564). Also, E.R.B. was supported by a Northern Forest Scholar fellowship from the Northeastern States Research Cooperative, a joint program of the University of Vermont, the University of Maine, and the Northern Research Station, USDA Forest Service.