Cumulative response of ecosystem carbon and nitrogen stocks to chronic CO2 exposure in a subtropical oak woodland

Summary Rising atmospheric carbon dioxide (CO2) could alter the carbon (C) and nitrogen (N) content of ecosystems, yet the magnitude of these effects are not well known. We examined C and N budgets of a subtropical woodland after 11 yr of exposure to elevated CO2. We used open-top chambers to manipulate CO2 during regrowth after fire, and measured C, N and tracer 15N in ecosystem components throughout the experiment. Elevated CO2 increased plant C and tended to increase plant N but did not significantly increase whole-system C or N. Elevated CO2 increased soil microbial activity and labile soil C, but more slowly cycling soil C pools tended to decline. Recovery of a long-term 15N tracer indicated that CO2 exposure increased N losses and altered N distribution, with no effect on N inputs. Increased plant C accrual was accompanied by higher soil microbial activity and increased C losses from soil, yielding no statistically detectable effect of elevated CO2 on net ecosystem C uptake. These findings challenge the treatment of terrestrial ecosystems responses to elevated CO2 in current biogeochemical models, where the effect of elevated CO2 on ecosystem C balance is described as enhanced photosynthesis and plant growth with decomposition as a first-order response.


Introduction
Many experiments have examined the responses of plant production and ecosystem carbon (C) balance to rising atmospheric CO 2 (Reich et al., 2006a;Norby & Zak, 2011). Results from these feature prominently in assessments of potential feedbacks between the biosphere and the changing atmosphere (Dolman et al., 2010). Compared to responses of photosynthesis and plant growth to elevated CO 2 , the response of soil C is less well understood, because changes in soil C content are difficult to detect (Smith, 2004). Increased C in soil in response to elevated CO 2 is sometimes found (Jastrow et al., 2005;Iversen et al., 2008Iversen et al., , 2012, although more frequently there is no effect, whether because of low statistical power or the absence of an important effect is unclear . Ecosystem-scale inventories assessing C balance responses to elevated CO 2 also often show no effect (Hungate et al., 1997b;Gielen et al., 2005;Gill et al., 2006;Niklaus & Falloon, 2006;Adair et al., 2009), although in aggregate some analyses suggest an effect is apparent (Luo et al., 2006). Thus, global models projecting future C dynamics of the biosphere have strong support for the effects of CO 2 on plant growth (Denman et al., 2007), but less empirical support for assumed effects on total ecosystem C storage. Our first goal in this work was to construct a complete C inventory for a subtropical oak woodland after 11 yr of exposure to elevated CO 2 , to test whether the CO 2 treatment altered total system C accumulation, and determine how any changes in C accumulation were distributed among plant and soil pools.
Total ecosystem C content is a function of plant growth and accumulation of plant biomass and detritus and also of C losses through microbial decomposition. Microbial decomposition is typically assumed to be a first-order process (Parton et al., 1987), responding predictably and constantly to changes in substrate supply, and thus is not expected to respond to elevated CO 2 independently of changes in substrate accumulation (Denman et al., 2007). Challenging this idea, inputs of C to soil can stimulate mineralization of native soil organic matter (Lohnis, 1926;Broadbent & Norman, 1947;Broadbent & Bartholomew, 1949;Van Veen et al., 1991), and increased atmospheric CO 2 has been shown to promote microbial activity (Dieleman et al., 2010) and even soil C loss (Hoosbeek, 2004;Trueman & Gonzalez-Meler, 2005;Carney et al., 2007;Hagedorn et al., 2008;Paterson et al., 2008;Taneva & Gonzalez-Meler, 2008;Langley et al., 2009;Trueman et al., 2009;Drake et al., 2011;Reid et al., 2012).
Thus, soil processes influence potential C accumulation in response to increasing atmospheric CO 2 , yet how and to what extent are not well understood. Our second goal in this work was to examine changes in soil microbial activity during the 11 yr of CO 2 enrichment, and to test whether patterns of CO 2 effects on soil microbial activity might help explain any effects (or lack of effects) of elevated CO 2 on soil C stocks.
Carbon cycling in ecosystems is linked to cycles of other elements (Finzi et al., 2011), such as nitrogen (N). Simulations of land carbon uptake using models with coupled N and C dynamics usually differ, and in many cases differ strongly, from those ignoring N (e.g. compare Cramer et al., 2001 andThornton et al., 2007), because N limits plant growth and C storage (LeBauer & Treseder, 2008), and because N cycling is sensitive to environmental change (Galloway et al., 2008). With N cycling included, simulations project smaller increases in terrestrial C storage in response to rising CO 2 , because N availability limits plant growth and its response to elevated CO 2 (Thornton et al., 2007;McMurtrie et al., 2008;Sokolov et al., 2008;Jain et al., 2009;Wang & Houlton, 2009;Friedlingstein & Prentice, 2010;Zaehle et al., 2010).
While model simulations bear out the importance of including N, these models do not necessarily demonstrate a consistent pattern of effect. Results differ in magnitude, direction and mechanism, suggesting that additional data and analyses are needed to evaluate conditions under which C-N coupling is important. For example, some simulations project only a modest limitation of terrestrial C uptake with coupled C-N interactions in the long term (at equilibrium), but strong effects of C-N interactions on the dynamics of C cycling and storage after disturbance (Gerber et al., 2010). Although the models generally agree that including N limitation of plant production reduces the terrestrial C sink, the magnitude of this effect is highly variable (Arneth et al., 2010). Experiments also indicate that C-N interactions are critical modulators of the long-term CO 2 fertilization response, but different experiments provide support for different mechanisms underlying that modulation. In some cases, C-N interactions appear to constrain strongly the CO 2 response (Reich et al., 2006a,b;Norby et al., 2010;Garten et al., 2011), but in others, plants appear able to access the extra N needed to support the growth response Drake et al., 2011). Effects of CO 2 concentration on microbial N transformations that influence the plant-soil distribution of N are extremely variable, with negative, positive and neutral effects observed for the same processes (D ıaz et al., 1993;Zak et al., 1993;Morgan et al., 1994;Zanetti et al., 1996;Hungate et al., 1997a,c;Johnson et al., 1997;Hofmockel & Schlesinger, 2007;van Groenigen et al., 2011van Groenigen et al., , 2012. Furthermore, other concomitant global environmental changes will modulate N constraints on C balance responses to elevated CO 2, including changes that alter N cycling directly, such as warming, altered precipitation and atmospheric N deposition, as well as indirect effects, such as changes in plant species composition. There is considerable debate as to the magnitude of the impact of such effects on ecosystem C sequestration, however (Jenkinson et al., 1999;Nadelhoffer et al., 1999;Arneth et al., 2010). Thus, both model simulations and data can be invoked to support N cycling constraining, increasing, or having little effect on the terrestrial C sink. Our third goal in this research was to compare C and N inventories in response to 11 yr of CO 2 exposure in a subtropical woodland, in order to test how rising CO 2 affects these elements in concert.
One of the challenges in investigating C-N interactions in ecosystem experiments is that the timescale of measurements of N cycling rates is typically far shorter than the timescale of N cycling processes that influence ecosystem responses. Elevated CO 2 can alter multiple processes within the soil N cycle simultaneously, with strong temporal dynamics, and with opposing impacts on plant N availability, making it very difficult to extrapolate short-term measurements to long-term effects. Following an isotope tracer over multiple years can help overcome this challenge. 15 N tracers reflect short-term effects on N cycling processes and integrate these into long-term effects on 15 N distribution among plant and soil components within the system. Because the 15 N is added in labile form, losses of added 15 N will be relatively larger than losses of total ecosystem N, so can be detected with greater sensitivity. Our fourth goal in this research was to use a long-term 15 N tracer to characterize changes in N distribution and N losses in response to elevated CO 2 .
Here, we report a whole system inventory of the C and N content of a scrub-oak ecosystem after 11 yr of experimental CO 2 exposure. We also show how soil microbial activity responded to chronic CO 2 exposure. We also report recovery and distribution of a 15 N tracer applied early in the experiment, in order to assess how elevated CO 2 alters the system-level distribution of labile N over the timescale of a decade.

Materials and Methods
The scrub-oak experiment occurred at the Merritt Island National Wildlife refuge on the east coast of Florida, USA (28°38′N, 80°42′W). After controlled burning, 16 open-top chambers were established over the regrowing vegetation, each covering 9.42 m 2 ground area, with 8 chambers receiving ambient air and 8 receiving ambient air + 350 ppm V CO 2 (referred to as the 'elevated CO 2 ' treatment). A large blower circulated air through each chamber at a rate of 24-30 m 3 min À1 , replacing the chamber air volume 1.3-1.6 times min À1 . The chambers increased air temperature and vapor pressure deficit while decreasing light , microenvironmental effects that did not significantly alter growth or species composition (Seiler et al., 2009). The experiment began in May 1996 and was maintained until June 2007.
In June-July 2007, all aboveground material was harvested from the chambers (see Seiler et al., 2009), and roots and soils were collected using multiple cores in each chamber (see Day et al., 2013). For aboveground biomass, all shoots were cut at the base of the stem, weighed immediately, and subsampled for the determination of water content and elemental analysis of leaves and stems. Ten surface cores (0-10 cm) and five deep cores were collected from each plot at 10 cm increments; all cores were 7 cm diameter. Core depth varied among plots from 2 to 3 m due to differences in the depth to the water

Research
New Phytologist horizon. For purposes of the element inventory conducted here, depth increments were combined into 0-10, 10-30, 30-60 and 60-100 cm. Samples were hand-picked to remove large roots, and subsamples separated into coarse particulate organic matter, roots and mineral soil. Belowground biomass was also sampled indirectly using ground-penetrating radar (Stover et al., 2007, Day et al., 2013. Material on the forest floor was gathered from 1/8th of each plot by hand, collecting until no visibly identifiable plant fragments remained. Material was dried, sifted to remove adhering sand, and weighed. We used a combination of density and biological fractionations to estimate soil carbon (C) pools of varying turnover rates. We used incubations to estimate labile and active soil C pools (and, by difference residual C), using the technique of Nadelhoffer (1990). We measured CO 2 production from laboratory incubations, combining short-term incubations of soils immediately after collection ) with 541-d incubations conducted in the lab at Northern Arizona University. We used density fractionations as described previously Carney et al., 2007), separating light (< 1.5 g cm À3 ), medium (1.5-1.8 g cm À3 ), heavy (1.8-2.2 g cm À3 ) and residual (> 2.2 g cm À3 ) organic matter fractions. Total soil C, N, 15 N and 13 C were also measured on bulk samples collected from the cores. Our fractionation analysis focused on soils from the 0-60 cm depths. For bulk soil analyses where we measure total C, N and 15 N, we present the data to 1 m to correspond with the depth of the root biomass inventory.
We measured microbial biomass using the chloroform-fumigation extraction method (Vance et al., 1987) in mineral soil (0-15 cm) sampled in July 1997; June, July, September and December 1998;September 1999;and May 2004. Soil subsamples (20-25 g at field moisture content) were extracted in 75 ml 0.5 M K 2 SO 4 before and after 24-h fumigation with ethanol-free chloroform. The K 2 SO 4 extracts were dehydrated in a forced-air drying oven at 60°C, the salts ground in a mortar and pestle, and the resulting powder analyzed for C, N, d 15 N and d 13 C on a CE 2100 elemental analyzer coupled to a Thermo DeltaPLUS-XL isotope-ratio mass spectrometer (http://www.isotope.nau.edu). Microbial biomass was calculated as the difference in mass (of C, N, 13 C or 15 N) between fumigated and nonfumigated samples, divided by 0.54 to correct for extraction efficiency (Vance et al., 1987). For samples collected after the 15 N tracer application (June 1998), we also measured the 15 N content of mineral soil (0-15 cm depth). After milling, soil N and 15 N contents were determined as described above.
The CO 2 added to the elevated-CO 2 treated plots was depleted in 13 C. We used a two-member mixing model to determine mineral soil C derived from new photosynthate (Leavitt et al., 1994;Hungate et al., 1996). Stem tissue produced in the elevated CO 2 treatment (d 13 C S,E ) provided an integrative measure of the d 13 C value of new photosynthate (average across five sampling dates, -42.6 AE 0.3 &). However, because mineral soil (d 13 C M,A ) and stem d 13 C (d 13 C S,A ) differed in the ambient C a treatment, we calculated the d 13 C signature of new carbon (d 13 C new ) as: The d 13 C of the mineral soil in the ambient CO 2 treatment was used as the end member for organic matter fixed before the experiment began. Carbon, N, 15 N, and 13 C were determined for all plant and soil components using coupled Dumas combustion isotope-ratio mass spectrometry (Carlo-Erba elemental analyzer and Finnigan Delta-V mass spectrometer) at the Colorado Plateau Stable Isotope Laboratory (www.isotope.nau.edu).
For testing soil microbial activity, we collected soil and litter samples in May through July of 2004, after 8 yr of CO 2 treatment. Soil sampling, preparation of microbial inocula, carbon and nutrient amendments, and incubation conditions are described in Brown et al. (2009). Carbon substrates included glucose and hot-water extracts of roots and leaf litter collected from the ambient and elevated CO 2 treatments. Microbial inocula from litter, rhizosphere and bulk soil communities were also prepared from the two CO 2 treatments. We used the BD-oxy system (BD Oxygen Biosensor System, BD Biosciences, Bedford, MA, USA (Garland et al., 2003;V€ ais€ anen et al., 2005;Zabaloy et al., 2008) to evaluate microbial respiration. The system uses a fluorophore that fluoresces as O 2 is consumed during the 48 h incubation. Normalized relative fluorescence was calculated as relative fluorescence after 48 h normalized by dividing by relative fluorescence after 1 h. The response to substrate addition was calculated as: Eqn 2 (R c , normalized relative fluorescence in the absence of resource addition; R r , normalized relative fluorescence with the added resource. Brown et al. (2009) present data from the ambient CO 2 treatment; here, we expand on this past analysis to evaluate responses of microbial respiration to elevated CO 2 . We used ANOVA to test for effects of habitat (rhizosphere, litter or bulk soil), inoculum source (ambient or elevated CO 2 ), substrate source (ambient or elevated CO 2 ), substrate type (litter or root), N, and P. We used a separate ANOVA to test compare responses to the addition of glucose vs natural substrates extracted from roots and litter. Where appropriate, ANOVAs were designed as split-plots, to account for the nonindependence of inocula collected from individual experimental plots subject to multiple combinations of resource treatments in the BD-Oxy assay. We used resampling to infer the effects and estimate the magnitude of the elevated CO 2 treatment on ecosystem C and N pools and recovery of tracer 15 N. We estimated 5% and 95% confidence limits for the difference in means between elevated and ambient CO 2 treatments, using 1000 samples with replacement (n = 8 for each treatment).

Results
Elevated CO 2 increased plant biomass, including the mass of C (g C m À2 ) in leaves, stems and coarse roots, and the total mass of C in plants (Table 1). The mass of C in fine roots was not significantly affected by the elevated CO 2 treatment at the final harvest (Table 1), although fine roots did exhibit significant increases at other times during the experiment (Day et al., 2013). On average, plant C accumulation by the end of the experiment was 71.5 g C m À2 yr À1 higher in elevated compared to ambient CO 2 , roughly equally distributed aboveground (37.5 g m À2 yr À1 ) and belowground (33.5 g m À2 yr À1 ). The C content of the litter layer, coarse particulate organic matter, total mineral soil C, and the light and medium density fractions did not significantly respond to the CO 2 treatment, whereas the heavy density soil C pool significantly declined. Elevated CO 2 had no effect on soil C in the spodic horizon, with no significant effect on total mineral soil C, or on the light, medium and heavy density fractions (Table 2); thus, C in the deep soil was also insensitive to the CO 2 treatment. In general, increased mass of plant C caused by elevated CO 2 did not translate to increased C storage in other ecosystem reservoirs (Table 1).
Elevated CO 2 increased the N content of plants aboveground (Table 3), but the N contents of coarse and fine roots did not respond to elevated CO 2 , yielding no effect on total plant N. The N content of most soil fractions was not significantly altered by elevated CO 2 , except the medium density fraction at 30-60 cm, which increased, and the light fraction at 10-30 cm, which declined. Increased C in plant pools with only small changes in N means higher C to N ratios. Higher C to N ratios under elevated CO 2 were observed for leaves, coarse roots and the sum of all plant parts; elevated CO 2 also increased the C to N ratio of the litter layer (Table 4). Elevated CO 2 did not increase the C to N ratio of any soil pool; the only soil pool to respondthe heavy density fractionactually declined in C to N ratio. Changes in plant and soil C to N ratios were compensatory, such that elevated CO 2 had no effect on the C to N ratio of the plant-soil system to 1 m depth.
Elevated CO 2 increased recovery of tracer 15 N in aboveground plant tissues, but reduced recovery in coarse roots, in the soil light fraction at 10-30 cm depth, and in the soil residual fraction at 0-60 cm (Table 5). Together, these changes resulted in a significant decline in whole-system 15 N recovery under elevated CO 2 . Elevated CO 2 reduced the d 15 N of plant tissue (weighted average of all plant parts), a dilution of the added 15 N tracer with unlabeled 15 N. This pattern indicates that elevated CO 2 increased plant access to N, either through new N inputs or redistribution from existing ecosystem N reservoirs. But, because total plant N did not respond to elevated CO 2 , the increase in inputs of new N to plants were matched by N losses from plants, such that CO 2 enhanced N turnover through the plant system. In contrast to plant d 15 N, the d 15 N of soils did

Research
New Phytologist not change with elevated CO 2 , nor was whole-system d 15 N affected (Table 6).
While elevated CO 2 did not alter total ecosystem C, and effects on soil C were either nil or negative, several results indicate that elevated CO 2 increased soil microbial activity. Elevated CO 2 increased C mineralization in laboratory incubations, particularly for the first 24 h after collection in the field (Fig. 1, and see McKinley et al., 2009), indicating a larger and more rapidly cycling labile soil C pool. Elevated CO 2 also increased the proportion of soil organic matter that occurred in the soil microbial biomass: averaged across seven sample dates from 1997 to 2004, more soil C, N and 15 N was contained in the soil microbial biomass in the elevated CO 2 treatment (P = 0.012 for C, P = 0.096 for N, and P = 0.049 for 15 N; Fig. 2). When common inocula were presented with the labile substrates produced by leaves and roots, substrates produced in the elevated CO 2 treatment were respired more completely than substrates from the same sources in the ambient CO 2 treatment (Fig. 3a), indicating that the substrates produced in the high-CO 2 environment were more susceptible to microbial decay. For the litter and rhizosphere microbial communities, microbial inocula from the elevated CO 2 treatment consumed more O 2 than inocula collected from the ambient CO 2 treatment when presented with a common C substrate (Fig. 3b). Glucose induced a greater response in bulk soil inoculum from the ambient treatment (Fig. 3b), which may reflect CO 2 -depletion of available soil C susceptible to priming (Brown et al., 2009).
The incorporation of the depleted d 13 C signature into organic matter pools revealed rates and patterns of flow of 'new' C into the system, where new C is that fixed since CO 2 fumigation a Values are means AE SE of the mean for the Ambient and Elevated CO 2 treatments, the Effect of the CO 2 treatment (E-A), and the bootstrapped 5% and 95% CLs (confidence limits) for the treatment effect. CPOM, coarse particulate organic matter. Soil fractions are density fractions, including light (<1.5 g cm À3 ), medium (1.5-1.8 g cm À3 ), heavy (1.8-2.2 g cm À3 ) and residual (calculated as total soil total minus the sum of measured density fractions). of new C, 31% of the total C contained in coarse roots (Fig. 4), yielding a mean C residence time in coarse roots of 35.5 AE 4.2 yr. The total difference in coarse root biomass between E and A was 480 g C m À2 . This could have been caused entirely by a stimulation of new root C (probably the most parsimonious interpretation), but it is possible that treatments differed in patterns of use of 'old', stored Can idea which should not be immediately dismissed, given that these plants use old C to build new roots (Langley et al., 2002). In the surface soil mineral fraction, the percent new C increased linearly (Fig. 5), with an overall mean residence time of C of 33.6 AE 2.1 yr. In the spodic horizon, there was no evidence of new C accumulation in the total mineral soil or in the light, medium, or heavy density fractions (Table 2). Overall, elevated CO 2 did not significantly alter the total C content of the system (Table 1), because increased C in plant reservoirs were compensated by reduced C from the soil (Fig. 6).

Discussion
In this subtropical oak woodland, 11 yr of exposure to elevated CO 2 increased plant C by 22%, with a smaller (and not significant) effect on plant N of 9%, well within the range of responses typically observed in plants growing under a wide variety of experimental conditions de Graaff et al., 2006;Luo et al., 2006). Absolute responses in the mass of C above-and belowground were similar, consistent with elevated CO 2 having little impact on the partitioning of biomass aboveand belowground (Tingey et al., 2000), in contrast to the expectation that root growth would increase disproportionately (Stulen & den Hertog 1993). In our experiment, the relative response aboveground was actually larger than that belowground, because most of the biomass in this system is belowground. The mean residence time of C in coarse roots (revealed by incorporation of the d 13 C tracer) was sufficiently long that, at the final harvest, only about one third of the C in coarse roots represented new growth over the course of this experiment. By contrast, all of the standing aboveground biomass at the final harvest had accumulated after fire. Thus, repeated cycles of fire disturbance and recovery might yield a larger cumulative response of new C in coarse roots. The increased C content of plants suggests the potential for elevated CO 2 to enhance ecosystem C uptake. Yet, increased C contained in plants was not reflected in the C content of soil, neither in the top meter nor in the deeper spodic horizon. Possibly, the experiment lacked sufficient power to detect soil C accumulation (Smith, 2004). Alternatively, other mechanisms may have operated to prevent soil C accumulation in this ecosystem. We can place boundary conditions on the power problem: integrated

Research
New Phytologist over the top meter of soil, the mean effect of CO 2 on total soil C was a decline of À44.3 g C m À2 yr À1 , with the 90% confidence interval spanning a range of CO 2 effects from more rapid losses of soil C (À132.4 g C m À2 yr À1 ) to gain (+ 57.9 g C m À2 yr À1 ). This range exhibits the power limitations typical when assessing responses of total soil C to elevated CO 2 . Isolating components of the total soil C reservoir can help overcome the problem of limited power (e.g. Iversen et al., 2012). In our case, we found that by year 6 of the experiment, elevated CO 2 had reduced the C contained in the light density (Carney et al., 2007) and in the acid-hydrolysable (Langley et al., 2009) fractions of soil C. These findings are consistent with the response we observed at the final harvest reported here where elevated CO 2 reduced the heavy density fraction of soil C (Table 1) and decreased soluble C susceptible to glucose-induced priming (Fig. 3). The pattern of declining soil C in soil fractions is difficult to reconcile with the concept of soil C accumulation as a first-order response to enhanced plant growth.
The second explanation for not finding soil C accumulation in response to elevated CO 2 is that it does not occur, because increased C input to soil is compensated by increased C loss. Elevated CO 2 could enhance export of C through leaching of dissolved organic matter. But, if elevated CO 2 increased leaching of C in this experiment, this response had no influence on the C content or d 13 C composition of the spodic horizon; the absence of any effect on d 13 C is especially unlikely if leaching was an important pathway for C loss. These findings indicate that elevated CO 2 did not substantially alter leaching losses of C from the system.
In contrast to the absence of any apparent effect on leaching, there was compelling evidence that elevated CO 2 increased the rate of C cycling through the soil: elevated CO 2 significantly increased the size and rate of C flow through the labile soil C pool (Fig. 1), it enhanced the proportion of soil C (and N, and 15 N) that were cycling through the soil microbial biomass (Fig. 2), and it increased the decomposability of labile plant substrates and promoted a physiologically more responsive microbial community (Fig. 3). Elevated CO 2 also increased fungal biomass, a Values are means AE SE of the mean for the Ambient and Elevated CO 2 treatments, the Effect of the CO 2 treatment (E-A), and the bootstrapped 5% and 95% CLs (confidence limits) for the treatment effect. CPOM, coarse particulate organic matter. Soil fractions are density fractions, including light (<1.5 g cm À3 ), medium (1.5-1.8 g cm À3 ), heavy (1.8-2.2 g cm À3 ) and residual (calculated as total soil total minus the sum of measured density fractions). as measured by ergosterol (Klamer et al., 2002), by direct measurements of mycorrhizal fungal biomass (Langley et al., 2003), and by the ratio of fungi to bacteria in the soil microbial biomass, as indicated by the analysis of phospholipid fatty acid profiles (Carney et al., 2007). These results indicate that higher microbial activity was associated with a shift in the composition of the microbial community. Increased soil microbial activity may also explain why the effect of elevated CO 2 on the C : N of plant tissues and the litter layer was not apparent, and indeed in some cases may even have been reversed, in soil organic matter. Specifically, elevated CO 2 increased the C : N ratio of individual plant tissues (Table 3) as commonly observed (Cotrufo et al., 1998;Norby et al., 2001), of the entire plant biomass, above-and belowground, and of the litter layer. Yet, this shift was not observed in soil organic matter after 11 yr of continuous inputs of plant material to the soil organic matter pool. There are two possibilities for this discrepancy: (1) either the inputs of plant material were too low compared to background soil organic matter to drive a change in soil organic C : N; or (2) by increasing soil microbial activity and the processing of C in the soil system, elevated CO 2 caused a compensatory response, tending to reduce soil C : N. Our finding that elevated CO 2 reduced the total mass of soil N in the medium density fraction, but increased it in the heavy fraction, is consistent with this second explanation. The medium fraction has a higher C : N ratio than the heavy fraction, and the medium fraction is thought to cycle into the heavy fraction as the soil organic matter is processed by microbial activity and interactions with minerals (Camberdella & Elliott 1992). Thus, the pattern we observe may indicate increased processing and turnover of soil N, promoting transfer to pools with lower C : N ratios, and a tendency for CO 2 to decrease soil C : N.
Some previous measurements at this site indicated that elevated CO 2 reduced or had no effect on microbial activity during the first 18 months of the experiment, with reduced gross N mineralization (Hungate et al., 1999) and either reduced or no impact on microbial biomass N (measured as ninhydrin-reactive N) and microbial activity (measured as fluorescein diacetate hydrolysis) in the rhizosphere (Schortemeyer et al., 2000), although the mechanism(s) for these changes were not apparent. These early responses were apparently transient, and did not indicate the decadal-scale response of soil microorganisms to elevated CO 2 . The measurements reported here of microbial biomass, the size of the labile soil C pool, and the distribution and retention of 15 N cycling through the system are more representative of the entire duration of the experiment (e.g. Fig. 2). Results from this experiment are consistent with the general finding that elevated CO 2 stimulates soil microbial activity (de Graaff et al., 2006;Dieleman et al., 2012), and the turnover of soil organic matter (Marhan et al., 2010;Phillips et al., 2012;Dawes et al., 2013).
Elevated CO 2 can stimulate microbial activity by increasing soil water content, especially in grasslands (Hungate et al., 1997a;Morgan et al., 2004), and this response can counterbalance the increased C inputs from enhanced plant growth at elevated CO 2 , causing no change in soil C accumulation (Marhan et al., 2010

Research
New Phytologist increased surface soil water content during the first several years , but this effect disappeared with leaf area development (Li et al., 2007), and elevated CO 2 had no effect on soil temperature . Thus, the changes in microbial activity and organic matter turnover that we observed are unlikely to have been driven by differences in temperature, although increased soil moisture may have played a role early on.
Elevated CO 2 can also increase microbial activity by enhancing the supply of C substrates to soil microorganisms, a response consistent with past reports that, in this experiment, elevated CO 2 stimulated the 'priming effect' (Carney et al., 2007;Langley et al., 2009), the phenomenon where there occurs 'extra decomposition of native soil organic matter in a soil receiving an organic amendment' (Bingeman et al., 1953). In the experiment described here, the O 2 consumption assay indicates that C derived from the litter and roots is more labile in the elevated CO 2 treatment (Fig. 3), leading to a larger quantity of labile organic matter (Fig. 1). The higher rates of microbial activity observed are consistent with the notion that these new inputs of labile C to soil increased mineralization of native soil organic matter (Van Veen et al., 1991;Carney et al., 2007). This phenomenon has been observed for some time (Lohnis, 1926;Broadbent & Norman, 1947;Broadbent, 1948) and evidence for it has grown: isotope tracer experiments in soil incubations show that substrate additions can more than treble the decomposition rate of native soil organic matter in the short term (Cheng & Johnson, 1998;Cheng et al., 2000). Substrate additions can influence the oxidation of old soil C reservoirs, for example, in deep soil (Fontaine et al., 2007), and can shape the response of soil C to elevated CO 2 (Hoosbeek, 2004;Trueman & Gonzalez-Meler, 2005;Carney et al., 2007;Hagedorn et al., 2008;Paterson et al., 2008;Taneva & Gonzalez-Meler, 2008;Langley et al., 2009;Trueman et al., 2009;Drake et al., 2011;Reid et al., 2012). Increased oxidation of old soil organic matter is likely a transient response to a change in the rate of labile C inputs. In the experiment described here, the reduction in soil C observed by year 6 (Carney et al., 2007)   www.newphytologist.com comparable to that found after 11 yr, suggesting that the substrates susceptible to priming-induced loss had mostly been degraded during the first 6 yr. The implications of this response are not limited to C: increased C input to soil, enhancing microbial activity and turnover, can also increasing nutrient availability to plants (Zak et al., 1993). Observations elsewhere that elevated CO 2 increases microbial activity in concert with greater plant N acquisition from soil are also consistent with this interpretation (Drake et al., 2011), although without direct evidence of increased soil organic matter turnover, increased root exploration is a simpler explanation. Results presented here call into question the notion that feedbacks stimulating soil microbial turnover and N availability necessarily lead to plant N accumulation and increased plant growth. On the one hand, we did find that elevated CO 2 stimulated plant N uptake and 15 N dilution in plant tissues, likely driven by increased turnover of soil organic matter mediated by microorganisms (Figs 1, 3; Johnson et al., 1998Johnson et al., , 2001Finzi et al., 2007). On the other hand, increased microbial activity

Research
New Phytologist likely promoted N losses, accounting for our finding that elevated CO 2 reduced recovery of added tracer 15 N (Table 4).
In this experiment, spanning more than a decade in a naturally occurring ecosystem, photosynthesis and aboveground plant growth exhibited strong responses to chronic exposure to elevated atmospheric CO 2 Seiler et al., 2009), leading to the increased aboveground C content reported here, as well as increased C in coarse roots (Day et al., 2013;Fig. 6). The elevated CO 2 treatment did not affect C in fine roots at the final harvest, although fine roots responded sporadically in this experiment, with particularly strong responses following the initial fire disturbance and after a hurricane in year 8 (Day et al., 2013). Elevated CO 2 did not increase soil C, and in fact tended to decrease it, likely a consequence of increased microbial activity. Elevated CO 2 also increased plant N uptake, possibly driven by higher microbial activity and increased soil N availability, but these responses were also associated with reduced recovery of a longterm 15 N tracer, likely indicating enhanced ecosystem N losses. Thus, CO 2 altered the C and N cycles in this ecosystem, but not in ways that promoted large or even detectable increments in total ecosystem C mass. The effect of elevated CO 2 on soil C turnover via the 'priming effect' was large enough to modulate net carbon balance. This finding is not unique, and treatment of this phenomenon in models of soil C cycling is likely warranted (Heimann & Reichstein, 2008;Chapin et al., 2009). While the importance of priming is becoming evident, the challenge to include the phenomenon in models is not trivial: priming is still poorly quantified and the mechanisms remain inscrutable. Meeting this challenge could improve substantially our understanding of terrestrial C cycling, replacing, or at least modifying, the stabilizing first-order kinetics of decomposition used in virtually all current models of the soil C cycle (Luo & Weng, 2011). The response of soil C to labile substrate inputs suggests a previously unrecognized sensitivity of what was thought to be a long-term, stable C sink in the biosphere.