Seasonal plasticity in the temperature sensitivity of microbial activity in three temperate forest soils

The temperature sensitivity of soil organic matter (SOM) decomposition has been a source of much debate, given the potential feedbacks with climate warming. Here, we evaluated possible seasonal variation in the temperature sensitivity of microbially mediated soil fluxes related to decomposition (net N mineralization, net nitrification, proteolysis, the maximum velocity (Vmax) of proteolysis, microbial respiration, and the Vmax of four soil exo-enzymes) across forests dominated by eastern hemlock (Tsuga canadensis), white ash (Fraxinus americana), and red oak (Quercus rubra) in central Massachusetts, USA. We asked two simple questions: (1) do temperature sensitivities vary across forest types or different steps of the decomposition process, and (2) do temperature sensitivities display plasticity on a seasonal time frame? We observed substantial variation in temperature sensitivities (Q10 and R10 values) across the different fluxes and forest types. The ash soils exhibited the strongest temperature sensitivities and the mineral-N fluxes exhibited higher temperature sensitivities relative to the proteolytic fluxes or microbial respiration. The Vmax of soil exo-enzymes varied considerably in an interactive manner across forests and time, and the response of some enzymes was consistent with the thermal plasticity. The enzymatic kinetic properties Vmax and Km (half-saturation constant) were strongly correlated with slopes that differed across enzymes, reflecting an enzyme-specific tradeoff between maximum catalytic rate and substrate-binding efficiency. Generally, Q10 values were largely constant, but R10 values varied in a manner consistent with distinct seasonal plasticity. There was a consistent seasonal shift in R10 values coincident with snowmelt, suggesting that the time following snowmelt is a particularly interesting and dynamic period of microbial activity in these temperate forests.


INTRODUCTION
The decomposition of organic materials in soils affects the global carbon (C) cycle by controlling nutrient supply for primary production and releasing CO 2 to the atmosphere (Attiwill and Adams 1993, Schlesinger and Andrews 2000, Raich et al. 2002, van der Heijden et al. 2008).The temperature sensitivity of decomposition has been an intensive and controversial topic of research given the potential feedbacks between climate warming and decomposition (Davidson et al. 2000, Giardina and Ryan 2000, Knorr et al. 2005, Davidson and Janssens 2006).It is clear that increasing temperature exponentially increases the rate of enzymatic reactions in well-mixed solutions, such that enzymatic reactions have an 'intrinsic' temperature sensitivity (Davidson and Janssens 2006).However, changes in substrate availability (i.e., adsorption and occlusion) and microbial activity (i.e., exo-enzyme production) can alter the observed or 'apparent' temperature sensitivity of decomposition (Davidson and Janssens 2006, Davidson et al. 2006, Subke and Bahn 2010, Conant et al. 2011).Understanding how various 'apparent' temperature sensitivities of decomposition emerge from a common set of underlying thermodynamic principles is a challenging area of current research (Kirschbaum 2010, Mahecha et al. 2010, Subke and Bahn 2010, Davidson et al. 2011).
Decomposition is not a single enzymatic step; it is a complex process of chemical transformations mediated by a suite of exo-enzymes, microbial consortia, and environmental factors.There is evidence that different steps in the decomposition process have different temperature sensitivities (Agren and Wetterstedt 2007, Bengtson and Bengtsson 2007, Balser and Wixon 2009).Conant et al. (2011) recently summarized the major steps of the decomposition process and their possible temperature dependencies.First, abiotic processes control the amount of soil organic matter (SOM) that is protected by physiochemical mechanisms (e.g., adsorption, occlusion).Second, microbial exo-enzymes depolymerize the unprotected, large, insoluble SOM compounds into soluble oligomers and monomers.Third, microbes assimilate these soluble compounds, use them for growth and metabolism, and release mineral forms as waste (e.g., CO 2 , NH 4 þ , NO 3 À ).The temperature dependence of each individual step may positively or negatively affect the overall rate of decomposition and its temperature dependence (Conant et al. 2011).While many studies have examined the temperature dependence of the final step of this process (CO 2 production and diffusion out of soil), there is much less information available on the temperature dependence of other decomposition steps (but see Macdonald et al. 1995, Knoepp and Swank 2002, Koch et al. 2007, Brzostek and Finzi 2012, Guntinas et al. 2012).
In addition to variation across the decomposition steps, there is also a debate regarding how the temperature sensitivity of decomposition may vary seasonally.Three hypotheses are supported by the literature.First, the simplest hypothesis is that a decomposition process has a constant temperature sensitivity, such that a kinetic rate parameter describing the temperature sensitivity of the process does not vary across seasons or temperatures (Fig. 1A-B).We refer to constant temperature sensitivity as the null hypothesis.Many models of decomposition and respiration assume this condition, and some observations and experiments support this simple hypothesis (Jones et al. 2005, Hartley et al. 2008, Malcolm et al. 2009, Mahecha et al. 2010).Second, there is evidence that thermal constraints on enzyme structure and microbial activity lead microbes to adapt evolutionarily or acclimate physiologically in response to seasonal changes in temperature.Often, this leads to processes having higher temperature sensitivities in colder environments, such that temperature sensitivity declines with increasing temperature (Tjoelker et al. 2001, Chen and Tian 2005, Bradford et al. 2008, Zhou et al. 2009, Brzostek and Finzi 2011).We refer to this as the thermal plasticity hypothesis (Fig. 1C-D); we intentionally use the term 'plasticity' to encompass a broad range of potential mechanisms, including physiological changes within individuals (acclimation) and evolutionary changes in gene allele frequencies (adaptation).Third, evidence largely from alpine meadows, but increasingly from other ecosystems such as temperate forests, suggest that processes related to decomposition exhibit distinct seasonal variation that is not directly and continuously related to temperature (Monson et al. 2006, Schimel et al. 2007, Schmidt et al. 2007, Schmidt et al. 2009, Kaiser et al. 2010, Kaiser et al. 2011).Processes exhibiting distinct seasonality are often related to changes in the composition of active microbial communities; fungi such as snow molds are highly active under snowpack, while bacteria dominate soil activity during the summer, and a distinct community is active during the transitional snowmelt period (Monson et al. 2006, Schmidt et al. 2007, McMahon et al. 2009, Schmidt et al. 2009).In this case, we expect the temperature sensitivity of a process to vary distinctly across seasons, without a directional pattern in response to soil temperature (Fig. 1E-F).We refer to this as the seasonal plasticity hypothesis.All three hypotheses are supported by literature, and it is not clear if different hypotheses are more applicable to particular ecosystem types or particular steps in the decomposition process.
We asked two simple questions: (1) do temperature sensitivities vary across forest types or the different steps of the decomposition process, and (2) are temperature sensitivities constant in time, or is there evidence of thermal or seasonally plasticity (i.e., which hypothesis is best supported, Fig. 1)?We investigated these questions by measuring the seasonal variation in apparent temperature sensitivities of five microbial-mediated soil fluxes across three types of temperate forests with widely varying soil C and N cycling characteristics.To further address the second question, we directly measured the kinetic properties of four soil exo-enzymes across a seasonal cycle.

Site
This study was performed in Harvard Forest, a Long-Term Ecological Research (LTER) site in central Massachusetts, USA (42.538 N,72.188 W).We used six replicate 8-m-radius plots of forest dominated by one of three target species assemblages: (1) eastern hemlock (Tsuga canadensis), (2) a mixture of red oak (Quercus rubra) and red maple (Acer rubrum), and (3) white ash (Fraxinus americana).We used the eastern hemlock and white ash plots that were established for a previous study (Brzostek and Finzi 2011) and identified six plots dominated by red oak with a minor red maple component for this study.We chose plots such that greater than 80% of the standing basal area was composed of the target tree species, the litter layer was dominated by leaves of the target species, and the central 5 m of the plot contained only the target tree species (Brzostek and Finzi 2011).All sites were located within a single contiguous forest tract (Prospect Hill, Harvard Forest), and the maximum distance between any of the 18 plots was ;1 km.All soils were Typic Dystrochrepts derived from glacial till, with granite-schist-gneiss bedrock.The three v www.esajournals.orgspecies assemblages were chosen to capture some of the biogeochemical diversity of temperate forests within a common climate condition.In particular, these forests represent a gradient in litter recalcitrance, strength of mycorrhizal associations, and organic vs. mineral nutrient use by the dominant trees, and thus represent a range of plant effects on soil nutrient cycling (Hobbie 1992, Chapman et al. 2006).
White ash trees produce litter with little lignin but high amounts of N, leading to a high rate of litter decomposition, little organic horizon development, and low soil C:N ratios (Melillo et al. 1982, Brzostek andFinzi 2011).White ash trees form arbuscular mycorrhizal fungal (AMF) associations and produce roots that have little direct effect on soil enzyme activities (Brzostek and Finzi 2011).
Eastern hemlock trees, in contrast, produce litter with high amounts of secondary compounds such as lignin and phenolics but relatively little N, leading to slow litter decomposition rates, the development of thick soil organic horizons, and the accumulation of soil organic matter (SOM) with a high C:N ratio in relatively acidic, base-poor soils (Finzi et al. 1998a, Finzi et al. 1998b, Ellison et al. 2005, Brzostek and Finzi 2011).Eastern hemlocks form ecto-mycorrhizal fungal (EMF) associations and produce roots that strongly increase soil enzyme activities, suggesting a greater importance of organic N cycling relative to the ash forests (Gallet-Budynek et al. 2009, Brzostek andFinzi 2011).
The forests co-dominated by red oak and red maple were chosen to represent an intermediate state between the eastern hemlock and white ash forests.These species produce litter that decomposes faster than hemlock but slower than ash litter, leading to intermediate SOM stocks, pH, and soil C:N ratios relative to ash or hemlock (Melillo et al. 1982, Finzi et al. 1998b).Red oaks form EMF associations and red maples form AMF, again placing these forests intermediate relative to the EMF hemlock and AMF ash forests (Berliner and Torrey 1989).This forest type is particularly important in this region, as forests co-dominated by red oak and red maple form the most common cover type across Harvard Forest and much of central Massachusetts (Foster et al. 1998, Urbanski et al. 2007).Reflecting the influence of the most dominant tree species, these categories are hereafter referred to as ash, oak, and hemlock soils.

Soil sampling over a seasonal cycle
We collected soils from the 18 sample plots across a seasonal cycle with particular focus on the spring period characterized by melting snow and increasing soil temperatures, given the importance of this time-period in other systems (Schmidt et al. 2007) and in this forest in particular (Brzostek and Finzi 2012).We sampled soil in the summer and fall of 2010 (June 9 and October 12), and under ;60 cm of snow in the winter of 2011 (March 1).We then sampled intensively during the snowmelt period of the subsequent spring-to-summer transition (April 11, May 3, and July 1), and continued sampling soil through the fall of 2011 (August 12, October 19).On each sampling date, three 4.5-cm diameter soil cores were collected from the top 15 cm of the mineral soil of each plot.The three cores were composited into a single sample per replicate plot, transported to Boston University in a cooler, and passed through a 2-mm brass sieve.Soil temperature at 10-cm depth was measured once per plot on each sampling date using a handheld type-K thermocouple (model HI9053; Hanna Instruments, Smithfield, Rhode Island, USA).

Temperature dependence of microbial activity
We measured the temperature sensitivity of five microbially mediated processes over a seasonal cycle by incubating sieved soil at field moisture conditions in temperature-controlled chambers at 4, 10, 17, 23, 30, and 358C.A wide temperature range was used to better resolve the response of microbial activity to temperature (Brzostek and Finzi 2012); note that soil temperature rarely exceeds 238C at this site (Savage andDavidson 2001, Savage et al. 2009).One subsample was measured per plot-replicate at each incubation temperature on each date, for .2500analyses.We measured five fluxes to capture a wide range of microbial activities, including mineral and organic cycling of N and C: (1) net N mineralization, (2) net nitrification, (3) proteolytic enzyme activity, (4) maximum proteolytic enzyme activity with saturating protein substrate, and (5) microbial respiration of CO 2 .
We measured the mineral N fluxes of potential net N mineralization and nitrification by incubating 20 g of sieved soil in 250-mL Nalgene containers for 28 days using standard methods (Finzi et al. 1998b, Brzostek andFinzi 2011).We measured the organic N fluxes of proteolytic rates and the maximum potential proteolytic rate with saturating substrate by incubating 2 g of sieved soil in 50-mL centrifuge tubes for 4 hours.Proteolysis, or the rate of protein depolymerization into free amino acids, was measured as the gross rate of extractable free amino acid production using the OPAME method as described elsewhere (Watanabe and Hayano 1995, Lipson et al. 1999, Jones et al. 2002, Brzostek and Finzi 2011).The V max of proteolysis was measured by adding saturating protein substrate to the incubation (0.6% casein in soil solution, as in Brzostek and Finzi 2011).To evaluate an integrated metric of microbial activity, we also measured the rate of soil CO 2 production (hereafter termed 'microbial respiration') by incubating 20 g of sieved soil in 500-mL glass jars.Headspace gas samples (10 mL) were taken three times at hourly intervals and injected into an infrared gas analyzer (Model EGM-4; PP-Systems, Amesbury, Massachusetts, USA) to determine the concentration of CO 2 .These concentrations increased at a linear rate for all samples (minimum r 2 ¼ 0.95, average r 2 ¼ 0.99).The rate of CO 2 production per g dry soil was calculated from these slopes, the jar volume, soil mass, and soil gravimetric water content.These measurements were performed 1 and 8 days after sieving, and the resulting data were averaged per subreplicate.All five fluxes were measured during the four sampling times surrounding the spring snowmelt.Proteolysis was measured at an additional time in the preceding fall (October 12), while net N mineralization and nitrification were measured in the preceding summer and fall (June 9 and October 12).
Temperature-response functions were fit to the data for the five measured fluxes for each replicate plot on each sampling date.We used the exponential Q 10 temperature-response function because the literature provides a basis for inferring plasticity (or 'acclimation') based on variation in the parameters of this equation (Atkin and Tjoelker 2003, Tjoelker et al. 2009, Finzi et al. 2011).The observed rate of a particular soil process (e.g., microbial respiration, proteolysis, etc.) was modeled as where R 10 is a pre-exponential coefficient and the rate at a reference temperature of 108C, Q 10 is the multiplicative change in rate per 108C change in temperature, and T is the incubation temperature (8C).Rather than fitting Eq. 1 directly, we fit a linear form of Eq. 1 using a natural-logarithm transformation The a and b parameters were estimated for each flux for each replicate plot on each sampling date by linear regression of Eq. 2 using the linear model function in R v.2.15.2 (R Development Core Team 2012).The Q 10 and R 10 parameters were calculated from a and b, respectively, as in Humphreys et al. ( 2005) The natural-log transformation in Eq. 2 stabilized the increasing variance with increasing rates that was observed in the raw data, and produced unbiased parameter estimates with normally distributed residuals.Different interpretations of the terms acclimation, adaptation, and sensitivity have been a source of confusion in the literature (Bradford et al. 2008, Hartley et al. 2009, Subke and Bahn 2010, Conant et al. 2011).
For the purposes of this project, we use the term 'plasticity' to describe statistically significant temporal variation in the fitted R 10 and Q 10 parameters.

Soil exo-enzyme kinetics
To further address the second question (see Introduction) we measured the seasonality of Michaelis-Menten kinetics for the activity of four exo-enzymes: acid phosphatase (AP) mineralizes phosphate groups from organic compounds, b-N-acetylglucosaminidase (NAG) decomposes microbial byproducts and cell wall constituents (amino sugars and chitin), and b-1,4-glucosidase (BG) and cellobiohydrolase (CBH) decompose cellulose.We did not measure the kinetics of oxidative enzymes that decompose recalcitrant humic materials such as lignin, as these enzymes use a free-radical reaction mechanism that does not conform to Michaelis-Menten kinetics (Claus 2004, Stone et al. 2011), although some studies have fit Michaelis-Menten equations to oxidativeenzyme activity data (Davidson et al. 2011, Wang et al. 2012).Exo-enzyme activities were measured on soil sampled from all 18 plots on five sampling dates in 2011 (March 1, April 11, May 31, August 12, and October 19), capturing a full seasonal cycle and a soil temperature range of ;1.5-198C.The maximum substrate-saturated exo-enzyme activities (V max values) were used to evaluate the hypothetical predictions across a seasonal cycle (Fig. 1).
We developed a microplate method to measure the Michaelis-Menten kinetics of the four exoenzymes based on previous methods utilizing fluorescent 4-methylumbelliferone (MUB) linked substrates (Finzi et al. 2006, Sinsabaugh et al. 2008, Stone et al. 2011).For each sample, the fluorescence of soil incubated with MUB-linked substrates was measured for each of 9 substrate concentrations in 4 subreplicate wells, soil background fluorescence was measured in 4 wells, the fluorescence of a MUB standard in the absence of soil was measured in 4 wells, the quenching of MUB fluorescence by the soil sample was measured in 4 wells, and the background fluorescence of the MUB-linked substrate was measured in 4 wells.Following a 1-hour pre-incubation period at 308C, the rate of increase in fluorescence was measured for each plate over a 30-minute period with a plate reader (SpectraMax Gemini XS, Molecular Devices, Sunnydale, California, USA).Raw fluorescence readings were corrected for background fluorescence and quenching.The slope of the increase in fluorescence with time was then converted to enzymatic activity (lmol substrate converted g dry soil À1 h À1 ) using the fluorescence standard readings and gravimetric soil water content.We fit Michaelis-Menten curves to the measured exoenzyme activities for each enzyme within each replicate plot on each date as where V max is the maximum exo-enzyme activity at saturating substrate concentrations, S is the substrate concentration, and K m is the halfsaturation constant, or the substrate concentration at which the exo-enzyme activity is half the V max value.Deviation from Michaelis-Menten behavior was investigated with Eadie-Hofstee diagrams, which plot a linear rearrangement of Eq. 5; no deviance was observed (Cervelli et al. 1973, Irving and Cosgrove 1976, Zivin and Waud 1982, Garcia et al. 1993).

Snow depth data
Daily average snow water equivalent (SWE) data were measured using a snow pillow located in a small clearing near the hemlock plots from late November 2010 through the end of May 2011 (Boose 2009).SWE was measured in this single location only and was used to illustrate temporal trends, not differences across species plots.We used a below-canopy camera at the Environmental Measurements Site (EMS) that is part of the Phenocam network (Richardson et al. 2007, Richardson 2008) to visually determine the date of snow arrival in the early winter (December 14, 2010) and the complete disappearance of snow in the spring (April 16, 2011).We only used SWE data from the snow pillow for the time-period between these dates.

Statistical analysis
We focused on understanding the variation in kinetic parameters derived from temperatureresponse curves (Q 10 and R 10 parameters) and Michaelis-Menten curves (V max and K m ) across three forest types (ash, oak, and hemlock soils) and 4-6 sampling dates.Data were analyzed with a repeated-measures completely randomized design (RM-CRD) considering the forest plot as the unit of replication (n ¼ 6).The MIXED procedure of the SAS software was used in all cases to analyze data with analysis of variance (ANOVA) considering species, date, and exoenzyme type as fixed effects (SAS 9.1; SAS Institute, Cary, North Carolina, USA).The covariance of repeated measures was modeled using an auto-regressive-1 framework and the denominator degrees of freedom were modeled using the Kenward-Roger procedure (Kenward and Roger 1997).Log-transformations were frequently required to meet the assumptions of analysis; log-transformations normalized and stabilized the variance of residuals, as residual variance tended to increase as flux rates increased.When significant main effects were observed, post-hoc tests were performed on all pair-wise comparisons using the Tukey adjustment to control the experiment-wise error rate.Post-hoc tests were not performed on the R 10 parameters across different fluxes, as these values have different units and thus cannot be meaningfully compared (e.g., it is not useful to statistically compare the R 10 values for net N mineralization vs. microbial respiration).Results were considered to be consistent with the null hypothesis of constant temperature sensitivity (Fig. 1A-B) when the main effect and interactions involving sampling date were not significant.Results with a significant effect of sampling date (main effect or interaction) and a significant correlation with soil temperature were considered consistent with thermal plasticity (Fig. 1C-D), while results with a significant date effect and no correlation with soil temperature were considered consistent with seasonal plasticity (Fig. 1E-F).

Temperature sensitivity
All five of the measured fluxes exhibited positive exponential temperature-dependent activity, with the sole exception of net nitrification in ash soils measured at 358C (Fig. 2B).The Q 10 function (Eq. 1) accurately described the temperature variation of measured fluxes.Q 10 functions fit to each plot-replicate explained 83-93% of the observed variation in fluxes with minimal bias (Appendix: Fig. A1, Table A1).
The Q 10 and R 10 parameters (Table 1) were generally higher in ash soils relative to oak and hemlock soils (Fig. 3; ANOVA with Tukey posthoc tests).The higher rates of mineral-N fluxes (net N mineralization and net nitrification) in ash soils (Fig. 2A-B) were driven by significantly higher Q 10 values (Fig. 3A), and by a significantly higher R 10 value for net nitrification relative to oak or hemlock soils (Fig. 3B).Similarly, the higher rates of organic-N cycling in ash soils relative to oak or hemlock soils (proteolysis and V max of proteolysis, Fig. 2C-D) were driven by significantly higher Q 10 and R 10 parameters (Fig. 3A-B).The higher rate of microbial respiration in ash relative to oak or hemlock soils (Fig. 2E) was explained by a higher R 10 value (Fig. 3B).
The observed temperature sensitivities also varied across the different fluxes.Across all soils, net N mineralization had relatively high Q 10 values and low R 10 values (Fig. 3), yielding temperature-response curves that were more strongly exponential than the other fluxes (Fig. v www.esajournals.org2).Proteolysis and microbial respiration, for example, had relatively low Q 10 values and high R 10 values (Fig. 3), yielding temperature-response relationships that were weakly exponential.The addition of saturating protein substrate increased the Q 10 of proteolysis from an average of ;1.4 to an average of ;1.7 for maximum proteolysis (Fig. 3A), suggesting that low substrate availability constrained the Q 10 values for proteolysis.
The R 10 values for all fluxes varied strongly across the sampling dates while Q 10 values were less variable in time, as the F-values of the date main effect tended to be smaller for Q 10 relative to R 10 values (Table 1).The Q 10 values for proteolysis and maximum proteolysis did not vary significantly in time (Table 1, Fig. 4), consistent with the null hypothesis (Fig. 1A-B).However, the Q 10 values for net N mineralization, net nitrification, and microbial respiration varied across sampling dates (Table 1), but in a manner that was not correlated with soil temperature (Fig. 4), consistent with the seasonal plasticity hypothesis (Fig. 1E-F).R 10 values also varied across seasons in a manner not directly related to soil temperature (Table 1, Fig. 5), consistent with the seasonal plasticity hypothesis (Fig. 1E-F).Only the R 10 of net nitrification had a significant date by species interactions (Table 1) and varied over time in ash soils but not in oak or hemlock soils, where nitrification rates were low (Fig. 2B).
The seasonal plasticity in R 10 values was driven by opposite responses of organic vs. mineral N fluxes following the spring snowmelt period (Fig. 5).Soils sampled in the winter beneath snowpack exhibited high proteolytic rates and maximum proteolytic rates at saturating substrate across all species (Fig. 5E, G).Following snowmelt, these rates declined for all species in April and May before recovering in June.In contrast, R 10 values for net N mineralization were maintained at high rates in ash soils and increased in oak and hemlock soils during the spring (Fig. 5A), at the same time that the R 10 values for organic N fluxes were declining (Fig. 5E, G).Thus, organic N fluxes declined following snowmelt, while mineral N fluxes increased.Microbial respiration R 10 values varied significantly over time (Table 1), but not in a predictable manner across species or dates (Fig. 5I).

Enzyme kinetics
The activity of exo-enzymes followed Michaelis-Menten substrate kinetics in the soil of all three tree species, including the activities of acid phosphatase (AP, Fig. 6A), b-1,4-glucosidase (BG, Fig. 6B), b-N-acetylglucosaminidase (NAG, Fig. 6C), and cellobiohydrolase (CBH, Fig. 6D).The substrate-dependence of all enzyme activities were well described by Michaelis-Menten equations, as Eadie-Hofstee plots for these enzymes were linear with no departures from normality (data not shown).Variation in the measured exoenzyme V max values was dominated by main effects of species and enzyme and statistically significant two-way interactions between species, Notes: Q 10 values reflect the multiplicative rate change per 108C change in temperature.R 10 values reflect the pre-exponential factor and the kinetic rate at a reference temperature of 108C.P values are adjusted to achieve an experiment-wise error rate of a ¼ 0.05 using the Bonferroni correction.Significant effects are in boldface.All analyses reflect repeated-measures ANOVA with six replicates for each of three dominant tree species.
The maximum reaction rates (V max ) varied across enzymes in the order of AP .. BG .. NAG .CBH (main effect of enzyme, Table 2), and this rank order did not vary in time, despite the significant interaction between enzyme and sampling date (Table 2).There was a significant interaction between sampling date and species (Table 2); in the spring and early summer, total exo-enzyme activity was higher in ash soils relative to oak and hemlock soils, however there were no species differences in total exo-enzyme activity in the late summer and fall (Tukey posthoc tests).The V max of some enzymes varied across species; the V max of cellulose-decomposing enzymes (BG and CBH) was higher in ash soils than oak or hemlock, and the V max of NAG was lower in oak soils than ash or hemlock (Figs.6-7, Species 3 Enzyme interaction, Table 2).
The species-by-date and enzyme-by-date interactions reflect some plasticity in enzymatic V max values (Table 2, Fig. 7).The V max of AP activity varied significantly across dates, but not in a manner that was correlated with soil temperature (Fig. 7A), reflecting seasonal plasticity (Fig. 1E-F).However, the V max of cellulosic enzymes in ash soils (BG and CBH) declined as soil temperatures increased (Fig. 7B, D), consistent with thermal plasticity (Fig. 1C-D).The V max of NAG activity was more variable, but tended to v www.esajournals.orgincrease with increasing temperatures in hemlock soils, reflecting thermal plasticity in the reverse direction of the cellulosic enzymes in ash soils.The half-saturation constant (K m ) results followed the V max results, as V max and K m were strongly correlated (Fig. 8).
The observed V max and K m values were significantly positively correlated, and the slope of this correlation varied across enzymes (Fig. 8).AP and BG generally had high V max values and moderate K m values, while CBH and NAG generally had low V max values but relatively high K m values.The slope of the relationship between V max and K m was significantly greater for AP and BG relative to CBH and NAG (ANCOVA, log-log plots, P , 0.001).We summarize these complex results in relation to the proposed hypotheses (Fig. 1) in Table 3.The null hypothesis of constant temperature sensitivity was consistent with (1) the Q 10 values for proteolysis, maximum proteolysis, (2) the net nitrification R 10 values in oak and hemlock soils, and (3) the V max of BG in oak and hemlock soils and the V max of NAG in ash and oak soils.The thermal plasticity hypothesis was only consistent with the V max observations of cellulosic enzymes in ash soils and the NAG V max in hemlock soils.The seasonal plasticity hypothesis was consistent with (1) the Q 10 observations for net N mineralization, net nitrification, and microbial respira-  (A, C, E, G, I) and R 10 vs. soil temperature (B, D, F, H, J) panels have the same scale.There were no significant correlations between R 10 values and soil temperature (B, D, F, H, J). tion, (2) all of the R 10 observations with the exception of net nitrification in oak and hemlock soils, the V max of AP in all soil types, and the V max of CBH in oak and hemlock soils.Thus, all hypotheses were supported to some degree.However, the Q 10 values exhibited less temporal variation than the R 10 or exo-enzyme V max values; Q 10 values varied by an average of 17% across sampling dates, while R 10 values varied by an average of 76% and the exo-enzyme V max values varied by an average of 185% (Table 4).Thus, the Q 10 observations showed relatively little variation across time, the R 10 values varied more strongly over time consistent with the seasonal plasticity hypothesis, and the exoenzyme V max values were highly variable and supported all three hypotheses.

DISCUSSION
We began this study with two questions: (1) do temperature sensitivities vary across forest types or the different steps of the decomposition process, and (2) do temperature sensitivities display thermal or seasonal plasticity?Generally the answer to the first question is yes; we observed substantial variation in temperature sensitivities (Q 10 and R 10 values) across the different fluxes and forest types (Figs.2-3, Table 1).The ash soils exhibited the strongest temperature sensitivities, and the mineral-N fluxes exhibited higher temperature sensitivities relative to the proteolytic fluxes or microbial  Notes: Enzymes included acid phosphatase, b-1,4-glucosidase, b-N-acetylglucosaminidase, and cellobiohydrolase.Maximum rates of enzyme activity (V max ) and the halfsaturation constant (K m ) were analyzed separately with repeated-measures ANOVA with six replicates for each of three dominant tree species measured on five sampling dates.Significant effects are in boldface.
v www.esajournals.orgFig. 7. Maximum potential exo-enzyme activity (V max ) measured across five sampling dates that differed in soil temperature.Enzyme names are abbreviated as in Fig. 6.Significant correlations were plotted as a solid line for hemlock forests and dashed lines for ash forests.Values reflect the mean of six replicate plots and error bars reflect 61 SE.Fig. 8. Michaelis-Menten kinetic parameters covaried across exo-enzymes.Enzyme names are abbreviated as in Fig. 6.Each value reflects an individual soil sample measured on one of the five sampling dates.The maximum substrate-saturated activity (V max ) was significantly linearly correlated with the half-saturation constant (K m ) for all enzymes.The slope of these correlations was significantly higher for AP and BG relative to CBH and NAG (ANCOVA, log-log plots, P , 0.001).v www.esajournals.orgrespiration (Fig. 3).The answer to the second question depends on the metric of temperature sensitivity and the flux in question.Generally, we observed seasonally variable R 10 values consistent with the seasonal plasticity hypothesis (Fig. 1E-F), less variable Q 10 values consistent with the null hypothesis of constant temperature sensitivity (Fig. 1A-B), and mixed results for exo-enzyme V max values (Table 3).There was a consistent seasonal shift in R 10 values coincident with snowmelt, suggesting that the spring transition is particularly important to the temperature sensitivity of decomposition in these forests.

Plasticity of temperature sensitivities
We observed seasonal variation in R 10 values with less variable Q 10 values (Tables 3-4), consistent with type II acclimation of temperature sensitivity (Atkin andTjoelker 2003, Finzi et al. 2011).Type I acclimation refers to changes in Q 10 parameters, while type II acclimation refers to changes in R 10 parameters (Atkin and Tjoelker 2003).In the plant literature, type II acclimation of respiration is thought to involve a dynamic adjustment of respiratory capacity (e.g., mitochondrial density) and substrate availability (e.g., soluble sugar concentrations) to reduce R 10 as ambient temperatures increase; this leads to a nearly homeostatic rate of respiration despite varying temperature (Atkin et al. 2005, Tjoelker et al. 2008, Crous et al. 2011).However, it is more difficult to attribute our observations of seasonal variation in R 10 values to particular mechanisms, as we measured the activity of the soil community as a whole.Thus, we choose to use the term 'plasticity' in place of 'acclimation' to encompass a wide range of potential mechanisms.It is possible that the observed seasonal plasticity of R 10 values (Fig. 5) could arise from seasonal variation in substrate availability, physiological variation within individual microbial cells, a change in the number of active microbial cells, a change in microbial community composition, or some combination of these or other mechanisms (Goddard and Bradford 2003, Bradford et al. 2008, Gershenson et al. 2009, Dijkstra et al. 2011).
The seasonal plasticity in R 10 values was most pronounced in the spring and was characterized by a decline in the R 10 of proteolysis and an increase in the R 10 of net N mineralization following spring snowmelt (Fig. 5A, E, G).That is, the activity of soil proteolytic enzymes declined strongly following snowmelt, while net N mineralization increased.These results are consistent with observations in alpine meadow ecosystems; following spring snowmelt, there is a crash in fungal biomass and the proteolytic enzymes they produce, which releases labile pools of amino acids and other nutrients, provides substrates for N mineralization and nitrification, and supports the growth of bacteria and plants (Schadt et al. 2003, Schmidt et al. 2007, Weintraub et al. 2007).These similar patterns of proteolytic activity and net N mineralization following snowmelt are remarkable, given that Notes: Support for each hypothesis is shown separately for each forest type: A refers to ash forests, O refers to oak forests, and H refers to hemlock forests.Exo-enzyme abbreviations are as follows: AP, acid phosphatase; BG, b-1,4-glucosidase; CBH, cellobiohydrolase; and NAG, b-N-acetylglucosaminidase. Exo-enzyme K m values are not presented here because of the strong correlation between K m and V max ; inferences concerning K m followed those of V max .
v www.esajournals.orgcold-temperate forests and alpine meadows differ markedly in vegetation type, climate, and soil development.Future research could investigate whether the soil microbial community in cold-temperate forests shifts from fungal dominance in winter to microbial dominance in summer; such a pattern would be consistent with the observed data and supported by literature demonstrating the importance of soil fungi and the enzymes they produce in cold forest soils (Uchida et al. 2005, Schmidt et al. 2009).
We observed high temporal variability and variable plasticity in the V max values of soil exoenzymes (Table 4, Fig. 7).The strongest pattern in these data is the decline in cellulose-degrading exo-enzyme V max values with increasing temperature in ash soils (Fig. 7B, D).In these ash forests, the fall leaf litter is generally consumed by May or June, possibly because of the combination of rapidly decomposing litter and soil mixing by earthworms at these sites (Brzostek and Finzi 2011).Thus, BG and CBH V max values may have declined with increasing soil temperature not because of temperature per se, but because microbes decreased enzyme production in re- Notes: All values were averaged across forest types for each sampling date: Minimum, minimum observed mean value; Maximum, maximum observed mean value; Percent change ¼ (maximum 3 minimum)/minimum 3 100.Exo-enzyme abbreviations are as in Table 3. Exo-enzyme K m values are not presented here because of the strong correlation between K m and V max .Flux abbreviations are: Net N min, net N mineralization; Net nit, net nitrification; Max.Proteolysis, maximum substratesaturated rate of proteolysis; Respiration, microbial respiration.v www.esajournals.orgsponse to seasonal declines in cellulosic substrates in the ash sites (Chro ´st 1991, Shackle et al. 2000, Allison and Vitousek 2005).

Exo-enzyme Michaelis-Menten kinetics
The activity of the four exo-enzymes conformed to Michaelis-Menten kinetics (Fig. 6).Additionally, the V max and K m values were strongly correlated (Fig. 8), such that soils samples containing enzymes with high maximum reaction velocities (high V max ) also had low enzyme binding affinities (high K m values).There is evidence for such an enzymatic tradeoff between maximum catalytic rate and substrate binding affinity across many types of enzymes, including soil exo-enzymes (Somero 1978, Hochachka and Somero 2002, Savir et al. 2010, Allison 2012, Stone et al. 2012).That is, for an enzyme to have a high maximum catalytic rate it must have a relatively high degree of conformational flexibility to allow for rapid delivery of ligands to the enzyme active site.This flexibility, however, broadens the distribution of possible enzyme conformational states, including conformations unsuited for binding substrates, which reduces the substrate binding affinity and increases the observed K m (reviewed by Hochachka and Somero 2002).Thus, the correlation between V max and K m and the marked differences across species plots (Table 2) suggest that some of the variation in V max observations was driven by variation in enzyme structure in addition to simple variation in enzyme quantity per unit soil.
We observed marked and sustained differences in the kinetic properties (slope of V max vs. K m ) across the four exo-enzymes studied here (Fig. 8).That is, at a given K m value, the V max of AP and BG was substantially higher than the activity of NAG or CBH.We speculate that this may be explained by the mobility of the products and substrates of these enzymes.The two enzymes with high V max /K m values (AP and BG) release simple monomeric products (phosphate and glucose, with molar masses of 95.0 and 180.2, respectively), while the two enzymes with low V max /K m values (NAG and CBH) release comparatively larger products (N-acetyl-b-D-glucosaminide and cellobiose, with molar masses of 221.2 and 342.3, respectively).Alternatively, the observed correlation between V max and K m and the variance in the slope of V max vs. K m may be explained by the nature of the enzyme assays and competitive inhibition of non-fluorescently labeled natural substrates.Wallenstein et al. (2011) pointed out that when the Michaelis-Menten equation (Eq.5) is applied to environmental samples, its assumptions no longer apply and the fitted V max and K m values no longer represent independent biochemical attributes of individual enzymes.Soil exo-enzyme activity assays measure the net activity of a population of similar enzymes that liberate a fluorescent tag from an exogenous, labeled substrate.Naturally occurring substrates also interact with the exoenzymes, but these interactions are not measured because the natural substrates lack the fluorescent tag.Thus the naturally occurring substrates operate like competitive inhibitors and increase the observed K m values (Wallenstein et al. 2011).We cannot exclude the possibility that this competitive inhibition by natural substrates was larger for NAG and CBH relative to BG and AP, as it is difficult to measure biologically meaningful substrate concentrations in the poorly mixed soil media.
The exo-enzyme kinetic results (Figs. 6-8) indicate that V max and K m values vary across forest types and co-vary across exo-enzymes, such that enzymatic models of decomposition should use separate parameters for different enzyme classes (Schimel andWeintraub 2003, Allison et al. 2010).This confirms a recent literature synthesis of exo-enzyme kinetic parameters, which largely relied on data from laboratory experiments with purified enzymes (Wang et al. 2012).Additionally, the exo-enzymes in our soil samples required a high concentration of substrate for the measured activity to saturate (.200 lM; Fig. 6).This concentration is higher than used in some previous studies (often ;40 lM; Saiya-Cork et al. 2002, Finzi et al. 2006, Weintraub et al. 2007), suggesting that future investigations of exo-enzyme V max should include preliminary measurements to determine the amount of substrate required to measure substrate-saturated activity.

Conclusions
All three of the proposed hypotheses were supported to some degree (Fig. 1, Table 3).However, the seasonal plasticity hypothesis was most commonly supported by the observed data (Table 3).This seasonal plasticity was observed most prominently in the R 10 values of decomposition fluxes.The most pronounced seasonal pattern was a decline in the capacity and activity of proteolytic enzymes and an increase in N mineralization following spring snowmelt, consistent with well-documented patterns in alpine meadows (Lipson et al. 1999).We also found that decomposition fluxes had a varying degree of temperature sensitivity; temperature response curves were nearly linear for some fluxes and strongly exponential for others (Figs.2-3).The V max of soil exo-enzymes varied considerably in an interactive manner across forests and time, and the response of some enzymes was consistent with the thermal plasticity hypothesis.Measured exo-enzyme kinetics reflected a tradeoff between V max and K m , but the magnitude of this tradeoff varied across exo-enzymes.Because many of these patterns were consistently observed across forests of varying biogeochemistry, including ash, oak, and hemlock soils, they may be applicable to many temperate forests.

Fig. 1 .
Fig. 1.Conceptual depiction of three hypotheses.The null hypothesis of constant temperature sensitivity predicts that temperature sensitivity is constant across seasons (A) and soil temperatures (B).The thermal plasticity hypothesis predicts that temperature sensitivity varies with soil temperature, leading to smooth transitions across seasons (C-D).The distinct seasonal plasticity hypothesis predicts that temperature sensitivity varies across seasons with sharp transitions (E) with no direct relationship with soil temperature (F).

Fig. 2 .
Fig. 2. The temperature response of five microbially mediated fluxes measured across three temperate forest types.Values reflect the mean of six replicate plots per species, averaged across 4-6 sampling dates, with soil incubated at six temperatures.Error bars reflect 61 SE of the six plots.Curves reflect exponential Q 10 functions fit to the data.Data for net nitrification of ash soils incubated at 358C were excluded from the curve fitting as outliers.

Fig. 3 .
Fig. 3.The distributions of kinetic parameters describing temperature sensitivity varied across forest types and fluxes.The proportional change in flux per 108C change in temperature (Q 10 ) and pre-exponential factor (R 10 ) parameters describe the temperature sensitivity of the measured fluxes.The center line of the box plot depicts the median, the edges of the box reflect the 25th and 75th percentiles, and the whiskers depict the 10th and 90th percentiles; extreme observations are shown as points.The letters across the bottom denote statistical analysis across forest types (Tukey post-hoc test), fluxes that share a letter did not differ significantly.R 10 values were only compared within individual fluxes.Net N Min, net N mineralization; Net Nit, net nitrification.

Fig. 4 .
Fig. 4. Variation in Q 10 parameters across seasons (A, C, E, G, I) and soil temperatures (B, D, F, H, J).Values reflect the mean of six replicate plots and error bars reflect 61 SE.The grey line shows the snowpack water equivalent as measured by a snow pillow.There were no significant correlations between Q 10 values and soil temperature (B, D, F, H, J).

Fig. 5 .
Fig. 5. Variation in R 10 parameters across seasons (A, C, E, G, I) and soil temperatures (B, D, F, H, J).Values reflect the mean of six replicate plots and error bars reflect 61 SE.The grey line shows the snowpack water equivalent as measured by a snow pillow.Units for R 10 values follow Fig. 2. Note that for each flux, the y-axes of the R 10 vs. time(A, C, E, G, I) and R 10 vs. soil temperature (B, D, F, H, J) panels have the same scale.There were no significant correlations between R 10 values and soil temperature (B, D, F, H, J).

Table 1 .
Statistical analysis of temperature kinetic parameters for five fluxes describing microbial-mediated N and C cycling in temperate forest soils.

Table 2 .
Statistical analysis of kinetic parameters of the soil exo-enzymes.

Table 3 .
Summary of support for the three hypotheses.

Table 4 .
Percent change in the measured parameters across sampling dates.