Revisiting the Hypothesis That Fungaltobacterial Dominance

Ecological Monographs, 85(3), 2015, pp. 457–472 Ó 2015 by the Ecological Society of America

Revisiting the hypothesis that fungal-to-bacterial dominance characterizes turnover of soil organic matter and nutrients

1,3 2 JOHANNES ROUSK AND SERITA D. FREY 1Section of Microbial Ecology, Department of Biology, Lund University, Lund, Sweden 2Department of Natural Resources and the Environment, University of New Hampshire, Durham, New Hampshire 03824 USA

Abstract. Resolving fungal and bacterial groups within the microbial decomposer community is thought to capture disparate life strategies for soil microbial decomposers, associating bacteria with an r-selected strategy for carbon (C) and nutrient use, and fungi with a K-selected strategy. Additionally, food-web model-based work has established a widely held belief that the bacterial decomposer pathway in soil supports high turnover rates of easily available substrates, while the slower fungal-dominated decomposition pathway supports the decomposition of more complex organic material, thus characterizing the biogeochemistry of the ecosystem. We used a field experiment, the Detritus Input and Removal Treatments, or DIRT, experiment (Harvard Forest Long-Term Ecological Research Site, USA) where litter and root inputs (control, no litter, double litter, or no tree roots) have been experimentally manipulated during 23 years, generating differences in soil C quality. We hypothesized (1) that d13C enrichment would decrease with higher soil C quality and that a higher C quality would favor bacterial decomposers, (2) that the C mineralized in fungal-dominated treatments would be of lower quality and also depleted in d13C relative to bacterial-dominated high-quality soil C treatments, and (3) that higher C mineralization along with higher gross N mineralization rates would occur in bacterial- dominated treatments compared with more fungal-dominated treatments. The DIRT treatments resulted in a gradient of soil C quality, as shown by up to 4.5-fold differences between the respiration per soil C between treatments. High-quality C benefited fungal dominance, in direct contrast with our hypothesis. Further, there was no difference between 13 the d CO2 produced by a fungal compared with a bacterial-dominated decomposer community. There were differences in C and N mineralization between DIRT treatments, but these were not related to the relative dominance of fungal and bacterial decomposers. Thus we find no support for the hypothesized differences between detrital food webs dominated by bacteria compared to those dominated by fungi. Rather, an association between high-quality soil C and fungi emerges from our results. Consequently there is a need to revise our basic understanding for microbial communities and the processes they regulate in soil. Key words: aboveground/belowground; biogeochemistry; carbon sequestration; decomposer ecology; food web; forest ecology; fungal and bacterial dominance; Harvard Forest Long-Term Ecological Research (LTER); soil C cycle.

INTRODUCTION incorporation of microbial dynamics into C cycle models is limited by insufficient knowledge of if and Soils contain vast and dynamic pools of carbon (C) how microbial community structure translates into that are fundamental for maintaining the balance of functional relevance (Fierer et al. 2007, Keiser et al. atmospheric CO concentrations (Denman et al. 2007). 2 2014). Most C models treat the microbial community Transformations of soil organic matter (SOM) mostly implicitly (e.g., Jenkinson et al. 1999), assuming that rely on the activity of soil decomposer microorganisms microbial activity is solely determined by abiotic factors, that control the release of greenhouse gases including such as temperature, moisture, and pH, while commu- CO2 (Waksman and Gerretsen 1931, Schlesinger and nity identity is not explicitly considered. This approach Andrews 2000), the storage of SOM and subsequent assumes that microbial communities, or their supported maintenance of soil physical structure and fertility (Le food webs, cannot be distinguished with regard to their Guillou et al. 2012), and nutrient regeneration essential functional impact in the soil system. Recent models have for plant nutrition (Waksman and Starkey 1932). started to explicitly consider the qualitative characteris- Despite their integral biogeochemical role, the explicit tics of decomposer microbial communities (phylogenic or functional composition; Fontaine et al. 2003, Manuscript received 20 September 2014; revised 22 January 2015; accepted 28 January 2015. Corresponding Editor: J. B. Fontaine and Barot 2005, Allison 2012) finding im- Yavitt. proved model performances (Wieder et al. 2013). While 3 E-mail: [email protected] merely resolving the microbial community into fungal 457 458 JOHANNES ROUSK AND SERITA D. FREY Ecological Monographs Vol. 85, No. 3 and bacterial components is a coarse division, there are discriminate predictably against different C isotopes, and some emerging general patterns to note. In fact, the plant C produced by C3 plants, for example, have a resolving fungal and bacterial decomposers has been well-constrained d13C signal (Hayes 2001). After incor- thought to capture disparate life strategies for soil poration into microbial biomass, the C is slightly microbial decomposers, associating bacteria with an r- enriched in 13C compared to the initial substrate. Thus, selected strategy for C and nutrient use, and fungi with a after a cycle of microbial use, the C released back to the K-selected strategy (Fontaine et al. 2003, Fontaine and soil environment as a result of microbial turnover has Barot 2005, Blagodatskaya and Kuzyakov 2008, Fierer become enriched in d13C (Abraham et al. 1998, Cifuentes et al. 2009). Moreover, food-web models predict very and Salata 2001). Since microbial derived C is of lower different outcomes with regard to C sequestration, quality compared to plant-derived C (A˚gren et al. 1996, nutrient turnover, and ecosystem stability depending Schiff et al. 1997, Cotrufo et al. 2013), the gradual on microbial community composition, especially in enrichment in d13C signal in SOC that develops with each terms of the relative dominance of fungi and bacteria microbial use cycle coincides with a gradual reduction in (Moore et al. 2004, Six et al. 2006, Neutel et al. 2007, SOC quality (Churchland et al. 2013). Although it is Bezemer et al. 2010). In general, a widely held theoretically possible that this relationship can be offset assumption is that bacterial decomposer pathways in by changes in the composition of compounds that make soil support high turnover rates of easily available up the SOC (e.g., an enrichment of lignin) due to substrates, while slower fungal-dominated decomposi- compound specific differences in d13C signals, the pattern tion pathways support the decomposition of more has been found to be robust within and between soils and complex organic material (Wardle et al. 2004, de Graaff to overshadow compound specific differences (Sollins et et al. 2010, de Vries et al. 2012). al. 2009, Gunina and Kuzyakov 2014). Consequently, The C-limited soil microbial community is dependent using differences in natural abundances of C isotopes, an on and regulated by the availability and quality of effective index presents itself from which we can infer the organic matter (Demoling et al. 2007, Egli 2010, Hobbie quality of C left in the soil, as well as that used by and Hobbie 2012, Inselbacher and Na¨sholm 2012). The microorganisms for both catabolism (release of CO2) and availability and quality of soil organic matter, in turn, is anabolism (incorporation into microbial biomarkers). regulated by plant inputs. The plant influence is To test how microbial communities were affected by provided both from aboveground sources such as litter SOC quality, an operational definition of SOC quality as input (Hieber and Gessner 2002, Romani et al. 2006, perceived by the microbial community is required. We Sayer et al. 2011, Rousk et al. 2013), and from used two indices to achieve this. First, a direct measure belowground sources, such as root exudates and of the microbial ability to use the SOC as measured deposition (Farrar et al. 2003). It has been demonstrated respiration per SOC. While this measure only includes that nutrient content (Henriksen and Breland 1999) and one aspect of the microbial C use (catabolism; not the C quality of litter (Rousk and Ba˚a˚th 2007) anabolism), it has been found to be a robust proxy for determines which components of the microbial commu- microbial assimilability of C for both growth and nity that respond. Similarly, the type and concentration respiration under stable-state laboratory conditions of root exudate additions to soil determine the (A˚gren et al. 1996, Schiff et al. 1997, Cotrufo et al. responsive fractions within the microbial community 2013). Second, the d13C enrichment can be used to track (de Graaff et al. 2010, Eilers et al. 2010, Reischke et al. the quality of C in SOC, microorganisms, or respired 2014, Rousk et al. 2015). Plant litter of high C quality CO2. appears to favor bacterial more than fungal growth, We used a long-term field experiment where plant while lower-quality plant litter additions shows the inputs have been manipulated in a temperate forest opposite relationship (Rousk and Ba˚a˚th 2007, Strick- ecosystem over 23 years. This Detritus Input and land et al. 2009). However, most experimental assess- Removal Treatments, or DIRT, experiment has exper- ments that have resolved microbial responses to date imentally adjusted litter input (no litter, double litter, or have involved the influence of single additions of litter or control) and tree root input (no tree roots) in a model root exudate compounds to soil, and most temperate, mixed-hardwood stand at the Harvard assessments have been done in laboratory systems. Forest Long-Term Ecological Research (LTER) site in Consequently, the effects of plant inputs on microbial Massachusetts, USA. This field experiment was com- responses have not yet been sufficiently validated in bined with (1) a natural abundance carbon isotope intact ecosystems with semi-continuous supplies of litter (d13C) study to quantify the plant input manipulation and root deposition. Moreover, how plant-input-in- effects on the quality of the SOC, and its contribution to duced microbial responses translate into differences in respired CO2 and incorporation into the microbial ecosystem function (e.g., C and N mineralization), community as determined using microbial biomarkers remains to be established. (phospholipid fatty acids, PLFAs); (2) growth rate and Stable isotope probing (SIP) presents a means by biomass based assessments of fungal and bacterial which the microbial use of C derived from plant material components resulting from the long-term plant-input of varying quality can be assessed. Biosynthetic processes manipulations; and (3) C (respiration) and N mineral- August 2015 SOIL C CONTROLS ON FUNGI AND BACTERIA 459

TABLE 1. Treatment codes and soil characteristics.

þ Soil moisture SOM SOC TN C:N NH4 þ Treatment, horizon Soil pH (% fm) (%) (mg C/g soil) (mg N g soil) (by mass) (lgNH4 -N/g soil) Double litter, 2X O 3.7 (0.1) 55.8 (4.0) 46.3 (4.8) 275 (20) 12.8 (1.4) 16.3 (2.7) 3.2 (1.0) M 4.2 (0.1) 32.2 (0.9) 13.0 (1.3) 67 (8) 3.7 (0.5) 15.6 (1.1) 3.8 (1.2) Control, Co O 3.8 (0.1) 52.3 (4.9) 42.9 (4.2) 248 (23) 11.7 (2.1) 16.3 (1.7) 5.6 (1.9) M 4.3 (0.3) 33.8 (0.7) 12.6 (0.7) 68 (4) 3.9 (0.4) 14.6 (2.0) 4.0 (1.7) No litter, 0X O 4.5 (0.2) 35.7 (2.6) 13.4 (4.6) 100 (13) 5.7 (0.6) 10.4 (5.6) 2.1 (0.3) M 4.4 (0.1) 31.0 (1.5) 10.3 (1.1) 47 (1) 3.0 (0.4) 15.9 (4.6) 3.2 (0.7) No OA, OA O 4.2 (0.1) 38.0 (1.5) 22.4 (1.0) 127 (9) 6.6 (0.5) 15.3 (0.7) 1.5 (0.2) M 4.6 (0.1) 19.6 (1.7) 4.6 (0.5) 19 (1) 1.6 (0.2) 12.7 (3.2) 2.4 (0.3) No roots, T O 4.0 (0.1) 51.0 (0.5) 33.0 (2.9) 192 (25) 9.5 (1.4) 15.9 (3.3) 1.8 (0.4) M 4.4 (0.1) 35.4 (0.8) 12.2 (0.9) 60 (10) 3.6 (1.1) 15.9 (2.7) 5.7 (2.2) No input, T0X O 4.4 (0.2) 39.3 (1.2) 18.4 (2.4) 95 (23) 5.2 (1.6) 16.6 (3.3) 4.8 (2.0) M 4.5 (0.0) 31.6 (1.3) 9.6 (0.4) 42 (2) 2.8 (0.4) 15.9 (3.6) 2.9 (0.8) Notes: Values are means (with SE in parentheses) of six (Co) or three (other treatments) field replicates in the O and M horizons. Soil pH is the KCl (0.1 mol/L) extract (1:5, mass : volume). Soil moisture was determined using gravimentric loss at 1058C for 24 h, as is given as mass percentage of fresh mass (fm). Soil organic matter (SOM) content was determined using loss on ignition (6008C þ for 16 h). Soil organic carbon (SOC) and total nitrogen (TN) were determined using an elemental analyzer. The NH4 concentration was determined using diffusion traps in a KCl extract (1 mol/L). See Methods for further details. ization (15N-pool dilutions) determinations to assess Canada mayflower (Maianthemum canadense Desf.) and microbial function. We addressed some long-standing a variety of ferns. The soil is derived from glacial till assumptions in soil microbial ecology by evaluating deposited (about 3 m deep) over granite, gneiss, and three main hypotheses. (H1) SOM formed by 23 years of schist bedrock, and is moderately well drained stony- increased plant litter inputs (e.g., double litter) should be loam classified as an Inceptisol (USDA taxonomy). The of higher quality and thus favor bacterial decomposers mean annual temperature is 78C and mean annual compared to SOM resulting from 23-years of reduced precipitation about 1100 mm. litter or root inputs (e.g., no litter, no tree roots), which The DIRT experimental plots were established at the will be enriched in lower-quality C and thus favor fungal site in 1990, as described in Bowden et al. (1993). The 13 decomposers. (H2) The C mineralized (d CO2) in the plots are 3 3 3 m, free of trees and saplings, and with low-plant-input treatments (and presumably dominated aboveground and belowground plant C inputs manip- 13 by low-quality C and fungi) will be enriched in d C ulated in a number of ways. The treatments include compared to the C mineralized from the high-C-input control (Co, normal annual aboveground litter inputs), treatments. (H3) Bacterial-dominated decomposition in double litter (2X, twice the litter inputs of the control high SOC quality treatments will have a higher plots), no litter (0X, annual aboveground litter inputs mineralization rate of both C and N and a higher excluded), no tree roots (T, plots trenched and root potential for nutrient leakage (gross N mineralization regrowth into plots prevented), no input (T0X, plots rates), and a lower potential to retain mineral N (N trenched and annual aboveground litter excluded), and assimilation rates) than the relatively more fungal- no OA (OA, soil O, and A horizons removed and dominated treatments. replaced with B horizon material from nearby excava- tions). The treatment codes are provided in Table 1. The METHODS plots (n ¼ 3) were set up in a stratified random approach Site, field experiment, and sampling at upslope, midslope, and downslope positions within The study site is located in Harvard Forest in north- the site (see Bowden et al. [1993] for more details), with central Massachusetts, and has previously been de- two randomly assigned control plots present in each of scribed in greater detail (Bowden et al. 1993). It is about the three blocks. a 1-ha, century-old, mixed-hardwood stand dominated The litter interception treatments (0X, T0X) were by northern red oak (Quercus borealis Michx. F.; 43% administered by placing plastic mesh shade cloth on the basal area), red maple (Acer rubrum L.; 19% basal area), plots from early September to late October (a period and paper birch (Betula papurifera Marsh.; 15% basal during which 95% of all litter fall occurs). After this area), with an herbaceous ground cover dominated by period, the intercepted litter was returned to the plots 460 JOHANNES ROUSK AND SERITA D. FREY Ecological Monographs Vol. 85, No. 3 with normal rates of litter input (Co, T, OA), or 5.74 TBq/mmol; Perkin Elmer, Waltham, Massachu- redistributed from the litter interception treatments (0X) setts, USA) and unlabeled Leu, resulting in 275 nmol/L to the double litter treatment (2X). In the treatments Leu in the bacterial suspension, and a 2 h incubation at without root input (T and T0X), 70–100 cm trenches, 228C without light. Fungal growth and biomass extending about 20 cm below the lower edge of the (ergosterol concentration) were determined using the rooting zone, were dug 0.5 m outside of the plot acetate-in-ergosterol incorporation method (Newell and boundaries. Corrugated fiberglass sheets (3 mm thick) Fallon 1991) adapted for soil (Ba˚a˚th 2001), with were installed to prevent roots from reestablishing, after modifications (Rousk and Ba˚a˚th 2007, Rousk et al. which the trenches were refilled. After noting signs of 2009), which estimates the rate of ergosterol synthesis as root reestablishment, trenching treatments were re- a measure of fungal growth. We used 1-[14C]acetic acid administered in 1999. This time, new trenches were (sodium salt, 2.04 GBq/mmol, 7.4 MBq/mL; Perkin dug down to 1 m depth. The original plastic barrier was Elmer) and unlabeled acetate resulting in a final acetate removed and replaced with a seven-ply laminate concentration of 220 lmol/L, and an incubation time of combining four layers of high density polyethylene and 4 h at 228C without light. three high-strength cord grids (Griffolyn; Reef Indus- Respiration was measured by transferring 1 (O tries, Texas, USA), folded over twice. The new dividers horizon) or 2 (M horizon) g soil into a 20-mL glass were extended 20–40 cm aboveground, and 30 cm below vial and purging the headspace with pressurized air. The the 1 m deep trenches. The OA treatments were vial was closed with a crimp lid and incubated for 24 h at established by removing the O and A horizons and 228C without light. Afterward, the headspace CO2 replacing them with B horizon soils excavated from concentration was analyzed using a gas chromatograph nearby. The treatments have been maintained continu- (GC) equipped with a methanizer and a flame ionization ously, and the no input (T0X) plots are kept free of detector (FID; YL6500 GC, YL instruments, Gyeonggi- seedlings and herbaceous vegetation and falling wood do, Korea). Sub-samples of the headspace (1.5 mL) were litter (more than twice each growing season). also transferred to He-purged gas-tight borosilicate The DIRT plots were sampled in late October 2013. tubes (12 mL; Labco, Lampeter, UK) for subsequent From each plot, four cores (8 cm wide 3 10 cm deep) isotope ratio measurements of the produced CO2. were separated into organic (O) and mineral (M) Microbial biomass was measured using the substrate horizons and then combined into composite samples induced respiration (SIR) method (Anderson and maintaining the two horizons and plot replication. The Domsch 1978). Glucose : talcum (4:1; 5 mg/g) was added treatment where mineral (B horizon) soil had replaced and mixed into the soil samples following the respiration the O and Oa horizons (treatment OA) at the field analyses. After 20 minutes, the atmosphere was purged experiment’s initiation had a poorly developed O of CO2 with pressurized air, after which the vials were horizon (;1–2 cm depth). The samples were sieved (4 again closed with crimp caps, and incubated for 2–3 h at mm O horizon; 2 mm M horizon), pebbles and roots 228C. The CO2 evolved was then determined as were removed, and then mixed and homogenized into described above. SIR-respiration was converted to press-seal polypropylene bags that were stored on ice in biomass using the relationship 1 mg CO2/h at 228C styrofoam insulated boxes during shipping. Microbial corresponds to 20 mg biomass C (recalculated from analyses were performed within five days of sampling. Anderson and Domsch 1978), and assuming a 45% C content in microbial biomass. Soil characterization The microbial phospholipid fatty acid (PLFA) com- Water content was determined gravimetrically (1058C position was determined using 1 (O horizon) or 2 g (M for 24 h), organic matter content by loss on ignition horizon) soil according to Frostega˚rd et al. (1993) with (6008C for 16 h), and soil pH in a 1:5 (mass : volume) modifications described by Nilsson et al. (2007). An extraction with 0.1 mol/L KCl. Total soil organic C and internal standard (methyl nonadecanoate fatty acid, N concentrations were determined by dry combustion 19:0) was added before the methylation step. The using a Flash 2000 elemental analyzer (Thermo Fisher derived fatty acid method esters were quantified on a Scientific, Bremen, Germany). Soil NH4 concentrations GC-FID as previously described (Frostega˚rd et al. were determined as part of the gross N mineralization 1993), and then analyzed for d13C content. The PLFAs determination. chosen to indicate bacterial biomass were i15:0, a15:0, i16:0, 16:1x9, 16:1x7c, 10Me16:0, cy17:0, i17:0, a17:0, Bacterial and fungal growth, biomass and PLFA 18:1x7, and cy19:0, while PLFA 18:2x6,9 was used to composition, and total respiration indicate fungal biomass (Frostega˚rd and Ba˚a˚th 1996). Bacterial growth was measured using the leucine Taxonomic groups were designated based on a sub- (Leu) incorporation method (Kirchman et al. 1985) selection from Ruess and Chamberlain (2010). adapted for soil (Ba˚a˚th et al. 2001) with modifications (Rousk et al. 2009), which estimates the rate of protein Gross N transformation rates synthesis as a measure of bacterial growth. We used a The N mineralization and consumption rates were mixture of radio-labeled Leu, [3H]Leu (37 MBq/mL, determined using the 15N pool-dilution method. Each August 2015 SOIL C CONTROLS ON FUNGI AND BACTERIA 461 homogenized soil sample was divided into to subsamples between C quality (respiration per unit SOC) and d13C (each 10 g fresh mass) in 200-mL, gas-tight, polypropyl- content of SOC and respired CO2 and N transformation ene cups, to which 350 lLNH4Cl (45 lg N/mL) enriched ratesweretestedusinglineartypeIImajoraxis to 98% 15N (Sigma Aldrich, St. Louis, Missouri, USA) regressions. The PLFA composition (molecular percent- were administered evenly using a pipette, followed by age of the 30 most abundant PLFAs) was analyzed with vigorous mixing with a stainless steel spatula to ensure a principal component analysis (PCA) after standardiz- even distribution, after which they were lidded. One set ing to unit variance. To directly test microbial responses of the subsamples was extracted using a 1 mol/L KCl to a C-quality effect, as measured with respiration per solution (1:5, mass : volume) 1 h after 15N addition, and unit SOC, type II major axis regressions were used. To the second set was treated identically after about 24 h conservatively estimate independence between field incubation at 228C without light. The amounts of treatments, these regressions were done on the means 14 þ 15 þ NH4 -N and NH4 -N were determined by isotope of each treatment, rather than individual field replicates ratio mass spectroscopy (IRMS) after diffusion to glass (i.e., assuming n ¼ 12 rather than n ¼ 42), to avoid false fiber traps, according to standard procedures (IAEA positives. 2001). The 14/15N-content of the glass fiber traps were dried at room temperature using a desiccator, transferred RESULTS to tin cups, and measured with a Flash 2000 elemental Soil characteristics analyzer coupled to a Delta V Plus via the ConFlow IV The SOM content was affected by both treatment interface (Thermo Fisher Scientific, Bremen, Germany) (F5,30 ¼ 14.0, P , 0.0001) and horizon (F1,30 ¼ 116, P , at the Stable Isotope Facility at the Department of 15 0.0001) with an interaction between the factors (F5,30 ¼ Biology, Lund University, Sweden. The N addition to 7.9, P , 0.0001). There was a higher concentration of these soils increased in the inorganic N concentration by SOM in the organic than in the mineral horizon, while approximately 1–2 lg N/g, corresponding to rather the treatments separated clearly between those that limited changes in mineral N concentrations, at ca. 50% þ received litter and those did not (Table 1). The SOC and increases in NH4 concentrations above initial levels. The N concentrations behaved very similarly to the SOM gross N transformation rates were estimated as described concentrations, leading to a stable C:N ratio, indistin- in Bengtson et al. (2005). guishable between horizons and treatments, at about 15 d13C measurements (Table 1). Soil moisture was also affected by both 13 12 treatment (F5,30 ¼ 8.45, P , 0.0001) and horizon (F1,30 ¼ All results on C/ C ratios are reported as per mil 69, P , 0.0001) with an interaction between the factors (%) difference using the Pee Dee Belemnite (PDB) (F5,30 ¼ 2.8, P ¼ 0.04; Table 1). The O horizon had standard in the ] notation: higher moisture than the M horizon, and the treatments with litter had higher moisture than those without litter d13Cð%Þ¼½ðR =R 1Þ 3 1000 sample std input (Table 1). However, the observed differences in 13 12 where Rsample is the C/ C ratio of the sample and Rstd soil moisture were related to variation in the water- is the 13C/12C ratio in the PDB standard. All IRMS holding capacity, driven by the SOM concentration, measurements were performed using various appliances rather than differences in microbial availability to water. þ coupled to a Delta V Plus via the ConFlow IV interface The NH4 concentration was not distinguishable be- (Thermo Fisher Scientific, Bremen, Germany) at the tween horizons or treatments, averaging about 3.4 lgN/ Stable Isotope Facility at the Department of Biology, g soil (Table 1). Soil pH was affected by treatment (F5,30 13 Lund University, Sweden. For measurements of d Cin ¼ 11.7, P , 0.001) and horizon (F1,30 ¼ 38.4, P , 0.0001) 13 soil a Flash 2000 elemental analyzer was used, for d C with an interaction between the two factors (F5,30 ¼ 4.0, in respired CO2 a GasBench II was used, and for the P ¼ 0.007). It was marginally higher in the M than O d13C in FAMEs derived from extracted PLFAs a Trace horizon, while the treatments without plant litter input GC Ultra gas chromatograph was used. The reported (0X, T0X) had marginal higher pH than those with litter d13C values for the PLFAs were corrected for the C input (Co, 2X; Table 1). atom added during the methylation step as in Williams 13 Fungal and bacterial growth et al. (2006). The d C in the respired CO2 was corrected 13 for the background signal of d CinCO2 in the sampled There were small but systematic effects of treatments headspace vials using standard procedures. and horizon, with an interaction between the factors for fungal growth per unit SOC (Fig. 1a, b). Higher litter Statistical analyses input appeared to stimulate fungal growth; however, a Treatment effects on all measured variables were clear effect was the high rate of fungal growth in the OA tested for using two-way analysis of variance (ANOVA) treatment, with fungal growth rates being higher overall with the field treatment as one factor (Co, 2X, 0X, T, in the O compared with the M horizon. The bacterial T0X, and OA) and soil horizon (O or M) as the other. growth per unit SOC was affected by treatment but not Treatment comparisons were done using Tukey’s pair- by horizon (Fig. 2c, d). Like for fungal growth, high wise comparisons at the a ¼ 0.05 level. The relationship rates of bacterial growth were observed in the OA 462 JOHANNES ROUSK AND SERITA D. FREY Ecological Monographs Vol. 85, No. 3

FIG. 1. (a, b) The fungal growth rate as determined using the acetate incorporation into ergosterol method (treatment F5,30 ¼ 12.4, P , 0.0001; horizon F1,30 ¼ 17.5, P , 0.0001; treatment 3 horizon F5,30 ¼ 3.0, P ¼ 0.02), (c, d) the bacterial growth rate as determined using the leucine incorporation into extracted bacteria method (treatment F5,30 ¼ 3.6, P ¼ 0.011; horizon F1,30 ¼ 0.64, P ¼ 0.43; treatment 3 horizon F5,30 ¼ 2.7, P ¼ 0.041), and (e, f) the logarithm of the ratio between fungal-to-bacterial growth (treatment F5,30 ¼ 3.3, P ¼ 0.018; horizon F1,30 ¼ 0.25, P ¼ 0.62; treatment 3 horizon F5,30 ¼ 1.2, P ¼ 0.33). The O (a, c, and e) and M (b, d, and f) horizon results are presented separately. Values represent the mean 6 SE of six (Co) or three (other treatments) ﬁeld replicates, and are presented per soil organic carbon (SOC). Statistics are from two-way ANOVAs. See Table 1 for treatment deﬁnitions.

treatment, but the treatments with litter additions (2X, suggesting a correlation between fungal dominance Co, T) tended to be only marginally lower than those and litter input. without (0X, T0X). This yielded a ratio between fungal Respiration and gross N transformation and bacterial growth that was affected by treatment but The respiration rate per unit SOC was affected by not horizon (Fig. 1e, f). Again, we observed a trend both treatment and horizon (Fig. 2a, b). Rates were where treatments with litter inputs (2X, Co, OA, T) were higher overall in the O compared to the M horizon and higher than those without litter input (0X, T0X) were particularly high in the OA treatment. However, August 2015 SOIL C CONTROLS ON FUNGI AND BACTERIA 463

FIG. 2. The microbial rates of (a and b) C (respiration; treatment F5,30 ¼ 10.7, P , 0.0001; horizon F1,30 ¼ 31.2, P , 0.0001; þ treatment 3 horizon F5,30 ¼ 0.99, P ¼ 0.49) and (c–f) N (NH4 production, panels c and d [treatment F5,30 ¼ 2.3, P ¼ 0.067; horizon F1,30 ¼ 8.1, P ¼ 0.0079; treatment3 horizon F5,30 ¼ 0.59, P ¼ 0.70], and consumption, panels e and f [treatment F5,30 ¼ 1.27, P ¼ 0.30; horizon F1,30 ¼ 0.02, P ¼ 0.86; treatment 3 horizon F5,30 ¼ 1.6, P ¼ 0.18]) transformation. The O (panels a, c, and e) and M (panels b, d, and f) results are presented separately. Values represent the mean 6 SE of six (Co) or three (other treatments) ﬁeld replicates, and are presented per soil organic carbon (SOC). Statistics are from two-way ANOVAs. there was also a tendency for higher respiration in signiﬁcant treatment or horizon differences were ob- þ treatments with litter input (2X, Co, OA, T) compared served for NH4 consumption per unit SOC estimates to those without litter input (0X, T0X); a trend observed (Fig. 2e, f). þ in the O but not in the M horizon. The NH4 production C and N mineralization relationships with SOC quality per unit SOC estimates gave a very similar, and yet clearer picture (Fig. 2c, d). Rates were higher in the O The d13C enrichment of SOC was distinguishable than M horizons, and treatments with litter input (2C, between horizons F1,30 ¼ 145, P , 0.0001) as well as

Co, OA, and T) were all clearly higher than those between treatments (F5,30 ¼ 6.4, P ¼ 0.0004) where an without litter input (0X and T0X) in the O horizon, with interaction (F5,30 ¼ 2.6, P ¼ 0.046) revealed larger no obvious treatment differences in the M horizon. No treatment effects in the O horizon. The SOC pool was 464 JOHANNES ROUSK AND SERITA D. FREY Ecological Monographs Vol. 85, No. 3

13 FIG. 3. The dependence of carbon (C) quality (as indicated by respiration per SOC) on the d C in (a) soil organic C (y ¼414 2 þ 16.1x; P ¼ 0.004, R ¼ 0.58) or (b) in respired CO2, and the dependence of nitrogen (N) quality (as indicated by NH4 production per SOC) on (c) the C quality (y ¼ 0.96x 10.5; P ¼ 0.0007, R2 ¼ 0.69) and (d) the d13C in soil organic C (y ¼15.4x 404; P ¼ 0.0003, R2 ¼ 0.74). The fitted relationship is a linear type II major axis regression, and is only shown when significant. The O and M horizons are shown as solid circles and open circles, respectively. Values represent the mean 6 SE of six (Co) or three (other treatments) field replicates.

more enriched in 13C in the M than the O horizon, and SOC (Fig. 3c) and by the d13C enrichment in SOC (Fig. arranged in order from least to most depleted as 0X (a), 3d). The slope between the rate of mineralization of C T0X (ab), Co (bc), T (bc), OA (bc), and 2X (c) per h and N per day (Fig. 3c) suggested that the C:N (treatments sharing the same lowercase letter are not ratio of the SOM mineralized was ;25, and thus higher signiﬁcantly different per Tukey’s test; Fig. 3a). This than the average C:N of the SOM (;15; Table 1). In 13 þ variation in d C enrichment of SOC correlated strongly contrast, the NH4 consumption rate did not show any with C quality, as indicated by respiration per unit SOC clear relationship with the quality of SOC. Instead, þ (Fig. 3a), where the most depleted SOC (OA and 2X strong positive relationships between NH4 consump- treatments in the O horizon with a d13C , 28.0%) had tion rates and both fungal (type II major axis regression: a fourfold higher quality than the most enriched SOC P ¼ 0.02, R2 ¼ 0.43, n ¼ 12), and particularly with (2X and T0X treatments in the M horizon with a d13C . bacterial growth (type II major axis regression: P ¼ 26.5%; Fig. 3a). In contrast, the ]13C enrichment in the 0.0006, R2 ¼ 0.71, n ¼ 12) could be substantiated. respired CO2 revealed no systematic horizon or treatment dependence, and thus C quality had no inﬂuence Microbial biomass 13 on the CO2 d C enrichment (Fig. 3b). The microbial biomass per unit SOC, as measured by þ The NH4 production showed a clear dependence on SIR, depended on treatment (F5,30 ¼ 10.1, P , 0.0001), the quality of SOC as indicated by both respiration per but only marginally on soil horizon (F1,30 ¼ 4.1, P ¼ August 2015 SOIL C CONTROLS ON FUNGI AND BACTERIA 465

TABLE 2. Soil microbial characteristics.

SIR-biomass Total PLFAs Bacterial PLFAs Fungal PLFA 18:2x6,9 Ergosterol Treatment (mg C/g SOC) (nmol/g SOC) (nmol/g SOC) (nmol/g SOC) (lg/g SOC) Double litter, 2X O 11.7 (0.73) 1493 (53) 560 (22) 203 (25) 253 (64) M 9.7 (0.38) 1708 (89) 766 (56) 93 (10) 129 (8) Control, Co O 13.1 (0.90) 1614 (75) 638 (24) 185 (12) 224 (29) M 9.5 (0.72) 1652 (96) 743 (41) 88 (11) 109 (11) No litter, 0X O 10.2 (1.16) 2011 (186) 768 (98) 74 (19) 111 (16) M 7.0 (0.81) 1353 (139) 569 (79) 40 (5) 73 (8) No OA, OA O 19.3 (1.26) 2072 (113) 717 (55) 293 (13) 361 (71) M 23.5 (11.49) 2283 (292) 938 (133) 121 (19) 132 (16) No roots, T O 12.1 (1.86) 1631 (271) 680 (119) 153 (36) 280 (16) M 7.5 (0.75) 1367 (175) 629 (81) 46 (6) 136 (26) No input, T0X O 13.7 (1.15) 2033 (317) 672 (40) 98 (2) 160 (9) M 8.9 (2.02) 1357 (140) 595 (38) 37 (3) 70 (6) Notes: Values are means (with SE in parentheses) of six (Co) or three (other treatments) ﬁeld replicates in the O and M horizons. Microbial variables are given per g soil organic carbon (SOC). SIR-biomass is the total microbial biomass as estimated using the substrate-induced respiration (SIR) method. Total PLFAs (phospholipid fatty acids) is a measure of the total microbial biomass, while bacterial PLFAs and fungal PLFA represent measures of the subcomponents bacteria and fungi. Ergosterol is a measure of total fungal biomass (see methods for further detail).

0.051). The OA treatment stood out with an approxi- without litter input, consistent with the fungal PLFA mately twofold higher SIR biomass than the other results. treatments, which were not distinguishable (Table 2). 13 The microbial biomass per unit SOC, as measured by Microbial PLFA composition and d C content total PLFA, depended on both treatment (F5,30 ¼ 4.4, P The ﬁrst principal component (PC1) of the PLFA ¼ 0.0042) and horizon (F1,30 ¼ 4.4, P ¼ 0.044), with an pattern explained a substantial fraction (30.9%) of the interaction showing a larger treatment inﬂuence in the variation in the data, while PC2 explained 13.8% (Fig.

M horizon (F5,30 ¼ 3.4, P ¼ 0.016). The total PLFA 4a). The C-quality effect on the PLFA composition concentration per unit SOC was higher in the O than M aligned closely along PC1, with high C-quality treat- horizon, and ranked from high to low as OA, T0X, 0X, ments aligning toward high PC1 values, and vice versa. Co,2X,andT(Table2).ThebacterialPLFA Thus, the O horizon samples clustered toward high PC1 concentration per unit SOC was unaffected by treatment values and M horizon samples toward low PC1 values. and horizon (Table 2). The fungal PLFA concentration The litter input treatments (2X, OA, Co) within the per unit SOC depended on both treatment (F5,30 ¼ 22.1, horizons clustered toward high PC1 values, with the P , 0.0001) and horizon (F1,30 ¼ 106, P , 0.0001), with opposite being true for treatments with no litter input an interaction (F5,30 ¼ 3.8, P ¼ 0.0083) showing that (0X, T0X) within each horizon. This led to signiﬁcant treatment effects were larger in the O horizon. effects of both treatment (F5,30 ¼ 19.8, P , 0.0001) and The fungal PLFA concentrations were higher in the O horizon (F1,30 ¼ 224, P , 0.0001) on PC1. horizon, and the concentration in the different treat- The d13C enrichment of PLFA markers did not ments ordered from high to low concentration as OA, depend on treatment or horizon (Fig. 5). There were 2X, Co, T, T0X, 0X, (Table 2). Thus fungal PLFA systematic differences between the enrichment of differ- concentration per unit SOC was higher with litter input ent PLFA markers (Fig. 5; one-way ANOVA; F18, 779 ¼ (OA, 2X, Co, T) than without (T0X, 0X). The 13.3, P , 0.0001), where the d13C enrichment ranged ergosterol concentration per unit SOC was affected by between relatively enriched markers such as 10Me18:0, both treatments (F5,30 ¼ 6.5, P ¼ 0.0003) and horizon 16:1x5, a17:0 and i14:0 at 26% to 27%, to relatively (F1,30 ¼ 47.9, P , 0.0001). The concentration was higher depleted markers, such as 18:1x7, cy19:0 and 18:2x6,9 in O than in M horizons, and the treatments ranked at 31% to 29% (Fig. 5). When aggregated into the from high to low concentrations as OA, T, 2X, Co, T0X, taxonomic categories listed in Fig. 5, there were

0X (Table 2). Thus, treatments with litter input had signiﬁcant differences between groups (F6,30 ¼ 14.4, P higher concentrations of ergosterol per SOC than those , 0.0001), with d13C enrichment ranging from highest to 466 JOHANNES ROUSK AND SERITA D. FREY Ecological Monographs Vol. 85, No. 3

FIG. 4. The microbial phospholipid fatty acids (PLFA) composition variation. The treatment variation can be seen in a plot of the scores of the two first components from a principal component analysis on (a) the microbial PLFA composition and (b) the markers that drive this sample separation can be seen in a loading plot of the two first components. The O and M horizons are shown as solid circles and open circles, respectively. Values represent the mean 6 SE of six (Co) or three (other treatments) field replicates. lowest as G/AMF, Actin, Gþ,G/F, F and G respiration per unit SOC) that had resulted from the (taxonomic groups defined in Fig. 5). field treatments. Even within the O horizon, the field treatments had resulted in threefold differences in SOC C-quality effects on microbial variables quality (respiration per unit SOC in OA compared with There was a significant influence on the fungal growth 0X) and, more generally, the treatments with litter (OA, per unit SOC by C-quality, as estimated by respiration 2X, Co, T) had resulted in SOC with on average a per SOC (Fig. 6a). In contrast, bacterial growth per unit twofold higher SOC-quality than that resulting from SOC (Fig. 6b) and the fungal to bacterial growth ratio treatments without litter input (0X, T0X; Fig. 2a). We (Fig. 6c) were both unaffected by C quality. The also note that the OA treatment (OA removal and biomass concentration of fungi (Fig. 7a, d) and bacteria replacement with B horizon soils from nearby) generated (Fig. 7b) per SOC along with the PLFA composition did a particularly high C quality. This result is consistent not significantly depend on the C quality. However, we with early observations a few years after the initiation of note clear positive trends with the C quality for this DIRT experiment (Bowden et al. 1993), along with concentrations of fungal PLFA 18:2x6,9 per unit SOC recent observations (Lajtha et al. 2014); however, the (Fig. 7a; P ¼ 0.06), ergosterol per unit SOC (Fig. 7c; P ¼ underlying mechanisms are neither clear nor possible to 0.16), and for a systematic variation in the PLFA evaluate in the current assessment. One possible reason composition (Fig. 7c; P ¼ 0.09), while the bacterial could be elevated pH during the past decades, but this biomass per unit SOC contrasted with this pattern by can at present only be substantiated with a marginally being highly independent of the C quality (Fig. 7b). elevated pH in the M horizon compared with other treatments (Table 1). Interesting follow-up questions DISCUSSION that emerge are whether the physical, chemical, or biotic DIRT treatment induced effects on SOC quality mechanisms cause these pronounced differences The two metrics used to assess SOC quality as (Schmidt et al. 2011). Specifically, it would be valuable perceived by the microbial decomposer community, to explore whether the chemistry of annual C inputs is respiration per SOC and d13C in SOM, appeared to be different between the OA and control or litter addition responsive to the field-experiment manipulations, and treatments, or if biophysical accessibility, related to the yielded corroborative results. Consequently, our hy- nature of C movement down the soil profile, was potheses could be addressed. We hypothesized (H1) that radically affected in the OA treatment. The treatments the decadal plant input manipulations of the DIRT where belowground plant C allocation was manipulated experiment would result in differences in the SOC by trenching, in contrast, did not clearly affect the SOC quality, and that relatively high-quality SOC would quality, suggesting that aboveground plant C inputs benefit bacteria relative to fungi, with the opposite being dominated the long-term effects on SOC quality in the true for relatively low-quality SOC. Our results clearly studied system relative to belowground plant C inputs. show that the first part of the hypothesis is supported Although the determination of mechanisms for a C- with 4.5-fold differences in C quality (as shown by quality gradient of SOM was not an objective of the August 2015 SOIL C CONTROLS ON FUNGI AND BACTERIA 467

13 FIG. 5. The d C in extracted PLFA markers in the (a) O and (b) M horizons. The treatments are color coded according to legend. Markers associated with specific microbial groups are noted as Gþ (gram positive bacteria), G (gram negative bacteria), Actin (actinomycetes), AMF (arbuscular mycorrhizal fungi), F (saprotrophic fungi), and Univ (universal). Values represent the mean 6 SE of six (Co) or three (other treatments) field replicates. present study, it is reassuring that our findings are mechanisms by which a gradient in SOC quality was consistent with previously reported results on litter- generated, we do see unambiguous evidence for one manipulation experiments based on SOC quality esti- (Fig. 2a, b). Added to this, both indices used to assess mations from a previous sampling of the same field SOC quality, the natural abundance of d13C in SOM experiment (estimated as time until 1% SOC loss by and respiration per SOC, provide corroborative evi- respiration in laboratory incubations; Lajtha et al. dence, thus matching previous findings (e.g., Abraham 2014). Additionally, although the belowground plant et al. 1998, Cifuentes and Salata 2001, Sollins et al. 2009, C input has been estimated to contribute substantially to Gunina and Kuzyakov 2014). We can consequently field measurements of soil respiration in DIRT experi- proceed to evaluate whether high-quality SOC favors ments, its contribution to SOM build-up has been bacteria, with the reverse being true for fungi (H1), and thought to be small in relation to the contribution by track the quality of used SOC with the d13C. litter (Sulzman et al. 2005). Moreover, in a recent assessment of a parallel DIRT experiment in deciduous Microbial responses to SOC quality forest stands in Oregon, root-derived C had a higher Fungal growth per unit SOC increased with higher resistance to decomposition than C derived from litter SOC quality (Fig. 6a), while both bacterial growth per (Crow et al. 2009). Irrespective of the underlying unit SOC (Fig. 6b) and the fungal-to-bacterial growth 468 JOHANNES ROUSK AND SERITA D. FREY Ecological Monographs Vol. 85, No. 3

ratio (Fig. 6c) showed no systematic pattern with soil C quality. While the biomass measures of fungi and bacteria did not show strong patterns with the SOC quality gradient resulting from the DIRT treatments, we note clear and reproducible suggestions for higher fungal biomass per unit SOC with higher-quality C (Fig. 7a, c), while bacterial biomass appeared completely independent (Fig. 7c). The d13C microbial PLFA was not affected by the DIRT treatments (Fig. 5). There were also very limited differences of d13C among most microbial PLFA markers (Fig. 5). In minor deviations from this general pattern, we note a trend for a use of particularly low-quality SOC by one of two actinomy- cete markers (i.e., enriched values in 10Me18:0), and a use of higher-quality SOC suggested by somewhat depleted values in the fungal marker (PLFA 18:2x6,9; Fig. 5). Added to this, the relative abundance of fungal PLFAs (PLFA 18:2x6,9) was higher in the high-C- quality treatments (Fig. 4). It should be noted that mycorrhizal fungal biomass normally is more depleted (by 2–3%) than saprotrophic fungal biomass (Ho¨gberg et al. 1999), which could also have contributed to both the relative abundance PLFA 18:2x6,9 and its d13C enrichment. However, together with the results from saprotrophic fungal growth, the d13C of fungal PLFA suggested use of higher-quality SOC than that used for most other PLFAs, further strengthening a consistent pattern where fungi were beneﬁted relative to bacteria by high-quality SOC. Thus, several lines of evidence add up to show that high-quality SOC beneﬁted fungal dominance over bacterial, in direct contrast with our hypothesis (H1).

SOC use by fungal and bacterial decomposers

We hypothesized that the CO2 produced in treatments that had resulted in lower-quality SOC with relatively higher dominance of fungi would have a more enriched d13C, indicative of more recalcitrant C being mineralized 13 (H2). In contrast, the d C in respired CO2 appeared to be independent of the SOC quality (Fig. 3d). This suggests that in treatments with a reduction in SOC quality (i.e., a lower relative abundance of high-quality C), less C was mineralized without any change in the quality of the C used. This result is consistent with results from previous assessments, which have found 13 13 similar d C in respired CO2 in spite of d C differences FIG. 6. The dependence of (a) fungal growth (y ¼ 8.1x þ in substrates respired (Fernandez et al. 2003, Church- 0.34; R2 0.38, P 0.034, F 6.01, h 12), (b) bacterial 13 ¼ ¼ ¼ ¼ land et al. 2013), and undistinguishable d CinCO2 growth, and (c) the fungal-to-bacteria growth ratio on carbon produced from different DIRT treatments in short term (C) quality (as indicated by respiration per SOC). The fitted relationship is a linear type II major axis regression, and is only assessments of fresh soil (Crow et al. 2006). Thus we see shown when significant. The O and M horizons are shown as evidence for a selective microbial use of high-quality solid circles and open circles, respectively. Values represent the components of the SOC, with a resulting respiration mean 6 SE of six (Co) or three (other treatments) field dominated by the abundance of high-quality C remain- replicates. ing, as suggested by the stable d13C (Fig. 3b) but variable rate (Fig. 2a, b) of respired CO2. As previously pointed out (e.g., Fernandez et al. 2003), the selective microbial processing of SOC with a relatively depleted d13C can also explain how the pattern between apparent August 2015 SOIL C CONTROLS ON FUNGI AND BACTERIA 469

FIG. 7. The dependence of (a) the fungal 18:2x6,9 concentration (b) bacterial PLFA concentration and (c) the PLFA composition (PC1 case scores), and (d) fungal ergosterol concentration on carbon (C) quality (as indicated by respiration per SOC). The O and M horizons are shown as solid circles and open circles, respectively. Values represent the mean 6 SE of six (Co) or three (other treatments) ﬁeld replicates.

SOC quality and its d13C content (Fig. 3a) has formed. differences. The rather higher C:N ratio of mineralized To evaluate our second hypothesis, we see no difference SOM (;25; Fig. 3c) relative to the C:N of the SOM between the SOC-quality used by a fungal compared (;15; Table 1) corroborates the evidence from the 13 with a bacterial-dominated decomposer community. depleted d C signal in the respired CO2 (Fig. 3b); both suggesting that plant-like material dominates the micro- Nutrient cycling of fungal and bacterial decomposers bial SOM use (Ho¨gberg et al. 1999, Gunina and We hypothesized that a relatively more bacterial- Kuzyakov 2014). þ dominated decomposer system should be characterized As has been previously noted, estimating NH4 by a nutrient cycle of lower retention, as shown by consumption with the presently used approach is biased þ higher C and N mineralization and a higher potential for by the increased NH4 -concentration that addition of mineral N leaching (Wardle et al. 2004, de Vries et al. 15N tracers induces. Thus, it likely results in overesti- 2012). Since mineralized organic matter is composed of mated consumption rates (Hart et al. 1994, Holub et al. both C and N, it should naturally be expected that C 2005), which could have compromised the sensitivity of þ mineralization and gross N mineralization should be NH4 consumption estimates to capture treatment well correlated (Hart et al. 1994). Our results conﬁrm differences (Fig. 2e, f). While no relationship between 2 þ this expectation (R ¼ 0.69, P ¼ 0.0007, n ¼ 12; Fig. 3c). NH4 consumption and any of our used SOC quality þ We observed treatment related differences in the NH4 indices could found, clear positive relationships were production, and thus the potential for mineral N seen to microbial growth; particularly to bacterial leaching, that were positively related with SOC quality. growth. So, did the detected differences in N transfor- This contrasted with previous assessments of other mation rates coincide with shifts in bacterial dominance? DIRT experiments where no treatment differences were There was no relationship of N or C mineralization with discerned (Holub et al. 2005). However, those previous measured bacterial variables (Figs. 1 and 2). Respiration assessments only included the mineral horizons, where (C mineralization) was closely related to fungal mea- we also observed less-pronounced treatment-induced sures, with a corroborative tendency also found for 470 JOHANNES ROUSK AND SERITA D. FREY Ecological Monographs Vol. 85, No. 3

þ NH4 production. Thus, the propensity for the ecosys- that these assumptions about the microbial processing tem to loose both C and N increased with a higher of SOM rarely have been tested, or, when they have, fungal dominance, in contrast with our hypothesis (H3). either the microbial use of SOM has been inferred from Moreover, the propensity for the system to retain the colonization of plant litter (e.g., de Graaff et al. þ mineral N (NH4 consumption rate) was more positively 2010) rather than SOM or biomass concentrations of related to bacterial than to fungal growth, again in fungi and bacteria have been assumed to reflect their contrast with the hypothesis (H3). contribution to SOM turnover (e.g., Wardle et al. 2004, de Vries et al. 2012), an assumption that is not met in Ecosystem implications of fungal-to-bacterial soil (Rousk and Ba˚a˚th 2007, 2011). dominance revised We find no evidence in favor of the supposition that To summarize and to evaluate our set of hypotheses bacterial-dominated food webs cycle faster and leak together, we found no support for a positive relationship more C and N than fungal-dominated ones, and between the relative dominance of bacteria and high convincing suggestions for the opposite, in the studied SOC quality (H1). The SOC decomposed in relatively forest soil. Our results highlight the fact that many of the more fungal-dominated soils is not different from that important dogma held as true in soil microbial ecology degraded in relatively more bacterial-dominated systems have limited empirical support (Strickland and Rousk (H2). The C and N mineralization, and thus the 2010). In one of the few empirical validations of another potential for C and N leakage, in relatively fungal- dogma, that fungi have a higher C-use efficiency than dominated systems was not lower than that occurring in bacteria (Six et al. 2006), Thiet et al. (2006) found that relatively bacterial-dominated ones; and fungal-domi- in fact the C-use efficiency of bacterial and fungal- nated systems were not better than bacterial-dominated dominated decomposer systems could not be distin- systems at retaining mineral N (H3). These results guished. Taken together with the results we report here, suggest structural differences between soil microbial the lack of support for central assumptions about the characteristics driven by the age and quality of C input, properties of the soil decomposer food web can only be that also carry implications for system biogeochemistry. regarded as humbling. Assumptions about the soil Specifically, increased fresh plant C influx will promote a decomposer community (e.g., C-use efficiency, nutrient shift toward a fungal dominance of SOM turnover. Such retention, etc.) influence predictions of ecosystem changes could be induced by increased forest produc- models (Allison 2012, Wieder et al. 2013) and the tivity (e.g., in warmer conditions) or induced by biomass empirical basis for these assumptions is still weak. harvest and subsequent recolonization (as indicated by ACKNOWLEDGMENTS the OA treatment). A shift toward fungal dominance will also lead to a faster mineralization of C and N, and We thank Mel Knorr for field, technical, and laboratory assistance, and Dr. Per Bengtson for stable isotopes method a somewhat lower microbial retention of mineral N. advice. This study was supported by the Swedish Research These findings contrast with widely held beliefs within Council (Vetenskapsra˚det) (grant no 621-2011-5719), the Lund soil microbial ecology. Detrital food webs dominated by University Infrastructure Grant, the Royal Physiographic bacteria are thought to support high turnover rates of Society of Lund, and by the U.S. National Science Foundation Long-Term Ecological Research (LTER) Program. easily available substrates, while slower fungal-dominated decomposition pathways support the decomposi- LITERATURE CITED tion of more complex organic material (Wardle et al. Abraham, W. R., C. Hesse, and O. Pelz. 1998. Ratios of carbon 2004, de Graaff et al. 2010, de Vries et al. 2012). The isotopes in microbial lipids as an indicator of substrate usage. crude phylogenetic resolution of fungal and bacterial Applied and Environmental Microbiology 64:4202–4209. ˚ decomposers has been thought to capture disparate life Agren, G., E. Bosatta, and J. Balesdent. 1996. Isotope discrimination during decomposition of organic matter: a strategies for soil decomposers, associating bacteria with theoretical analysis. Soil Science Society of America Journal an r-selected strategy for C and nutrient use, and fungi 60:1121–1126. with a K-selected strategy (Fontaine et al. 2003, Allison, S. D. 2012. A trait-based approach for modelling Fontaine and Barot 2005, Fierer et al. 2009). These microbial litter decomposition. Ecology Letters 15:1058– 1070. assumptions have been incorporated into food-web Anderson, J. P. E., and K. H. Domsch. 1978. A physiological models, where the different properties of fungi and method for quantitative measurement of microbial biomass bacteria with regard to C sequestration, nutrient in soils. Soil Biology & Biochemistry 10:215–221. Ba˚a˚th, E. 2001. Estimation of fungal growth rates in soil using turnover, and ecosystem stability have been concluded 14 (Moore et al. 2004, Six et al. 2006, Neutel et al. 2007, C-acetate incorporation into ergosterol. Soil Biology & Biochemistry 33:2011–2018. Bezemer et al. 2010). This influential work has Ba˚a˚th, E., M. Pettersson, and K. H. So¨derberg. 2001. manifested a widely held belief that bacterial decom- Adaptation of a rapid and economical microcentrifugation poser pathways in soil support high turnover rates of method to measure thymidine and leucine incorporation by easily available substrates, while slower fungal-domi- soil bacteria. Soil Biology & Biochemistry 33:1571–1574. Bengtson P, U. Falkengren-Grerup, and G. Bengtsson. 2005. nated decomposition pathways support the decomposi- Relieving substrate limitation—soil moisture and tempera- tion of more complex organic material (Wardle et al. ture determine gross N transformation rates. Oikos 111:81– 2004, de Graaff et al. 2010, de Vries et al. 2012). We note 90. August 2015 SOIL C CONTROLS ON FUNGI AND BACTERIA 471

Bezemer, T. M., et al. 2010. Divergent composition but similar Fierer, N., M. S. Strickland, D. Liptzin, M. A. Bradford, and function of soil food webs of individual plants: plant species C. C. Cleveland. 2009. Global patterns in belowground and community effects. Ecology 91:3027–3036. communities. Ecology Letters 12:1238–1249. Blagodatskaya, E., and Y. Kuzyakov. 2008. Mechanisms of Fontaine, S., and S. Barot. 2005. Size and functional diversity real and apparent priming effects and their dependence on of microbe populations control plant persistence and carbon soil microbial biomass and community structure: critical accumulation. Ecology Letters 8:1075–1087. review. Biology and Fertility of Soils 45:115–131. Fontaine, S., A. Mariotti, and L. Abbadie. 2003. The priming Bowden, R. D., K. J. Nadelhoffer, R. D. Boone, J. M. Meillo, effect of soil organic matter: a question of microbial and J. B. Garrison. 1993. Contributions of aboveground competition? Soil Biology & Biochemistry 35:837–843. litter, belowground litter, and root respiration to total soil Frostega˚rd, A˚., and E. Ba˚a˚th. 1996. The use of phospholipid respiration in a temperate mixed hardwood forest. Canadian fatty acid analysis to estimate bacterial and fungal biomass in Journal of Forest Research 23:1402–1407. soil. Biology and Fertility of Soils 22:59–65. Churchland, C., S. J. Grayston, and P. Bengtson. 2013. Spatial Frostega˚rd, A˚., A. Tunlid, and E. Ba˚a˚th. 1993. Phospholipid variability of soil fungal and bacterial abundance: conse- fatty-acid composition, biomass, and activity of microbial quences for carbon turnover along a transition from a communities from 2 soil types experimentally exposed to forested to clear-cut site. Soil Biology & Biochemistry 63:5– different heavy-metals. Applied and Environmental Micro- 13. biology 59:3605–3617. Cifuentes, L. A., and G. C. Salata. 2001. Significance of carbon Gunina, A., and Y. Kuzyakov. 2014. Pathways of litter C by formation of aggregates and SOM density fractions: Impli- isotope discrimination between bulk carbon and extracted 13 phospholipid fatty acids in selected terrestrial and marine cations from C natural abundance. Soil Biology & environments. Organic Geochemistry 32:613–621. Biochemistry 71:95–104. Cotrufo, M. F., M. D. Wallenstein, C. M. Boot, K. Denef, and Hart, S. C., G. E. Nason, D. D. Myrold, and D. A. Perry. 1994. E. Paul. 2013. The Microbial Efficiency-Matrix Stabilization Dynamics of gross nitrogen transformation in an old-growth (MEMS) framework integrates plant litter decomposition forest: the carbon connection. Ecology 75:880–891. with soil organic matter stabilization: Do labile plant inputs Hayes, J. M. 2001. Fractionation of carbon and hydrogen form stable soil organic matter? Global Change Biology isotopes in biosynthetic processes. Stable Isotope Geochem- 19:988–995. istry 16:1115–1128. Crow, S. E., K. Lajtha, T. R. Filley, C. W. Swanson, R. D. Henriksen, T. M., and T. A. Breland. 1999. Evaluation of Bowden, and B. A. Caldwell. 2009. Sources of plant-derived criteria for describing crop residue degradability in a model carbon and stability of organic matter in soil: implications for of carbon and nitrogen turnover in soil. Soil Biology & global change. Global Change Biology 15:2003–2019. Biochemistry 31:1135–1149. Hieber, M., and M. O. Gessner. 2002. Contribution of stream Crow, S. E., E. W. Sulzman, W. D. Rugh, R. D. Bowden, and detrivores, fungi, and bacteria to leaf breakdown based on K. Lajtha. 2006. Isotopic analysis of respired CO during 2 biomass estimates. Ecology 83:1026–1038. decomposition of separated soil organic matter pools. Soil Hobbie, J. E., and E. A. Hobbie. 2012. Amino acid cycling in Biology & Biochemistry 38:3279–3291. plankton and soil microbes studied with radioisotopes: de Graaff, M.-A., A. T. Classen, H. F. Castro, and C. W. measured amino acids in soil do not reflect bioavailability. Schadt. 2010. Labile soil carbon inputs mediate the soil Biogeochemistry 107:339–360. microbial community composition and plant residue decom- Ho¨gberg, P., A. G. Plamboeck, A. F. S. Taylor, and P. M. A. position rates. New Phytologist 188:1055–1064. Fransson. 1999. Natural 13C abundance reveals trophic de Vries, F. T., J. Bloem, H. Quirk, C. J. Stevens, R. Bol, and status of fungi and host-origin of carbon in mycorrhizal R. D. Bardgett. 2012. Extensive management promotes plant fungi in mixed forests. Proceedings of the National Academy and microbial nitrogen retention in temperate grassland. of Sciences USA 96:8534–8539. PLoS ONE 7:e51201. Holub, S. M., K. Lajtha, J. D. H. Spears, J. A. To´th, S. E. Demoling, F., D. Figueroa, and E. Ba˚a˚th. 2007. Comparison of Crow, B. A. Caldwell, M. Papp, and P. T. Nagy. 2005. factors limiting bacterial growth in different soils. Soil Organic matter manipulations have little effect on gross and Biology & Biochemistry 39:2485–2495. net nitrogen transformations in two temperate forest mineral Denman, K. L., et al. 2007. Couplings between changes in the soils in the USA and central Europe. Forest Ecology and climate system and biogeochemistry. Pages 663–754 in S. Management 214:320–330. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. IAEA. 2001. Use of isotope and radiation methods in soil and Averyt, M. Tignor and H .L. Miller, editors. Climate change water management and crop nutrition. IAEA Training 2007: the physical science basis. Contribution of Working Course Series No. 14. International Atomic Energy Agency, Group I to the Fourth Assessment Report of the Intergov- Vienna, Austria. ernmental Panel on Climate Change. Cambridge University Inselbacher, E., and T. Na¨sholm. 2012. The below-ground Press, Cambridge, UK. perspective of forest plants: soil provides mainly organic Egli, T. 2010. How to live at very low substrate concentrations. nitrogen for plants and mycorrhizal fungi. New Phytologist Water Research 44:4826–4837. 195:329–334. Eilers, K. G., C. L. Lauber, R. Knight, and N. Fierer. 2010. Jenkinson, D. S., J. Meredith, J. I. Kinyamario, G. P. Warren, Shifts in bacterial community structure associated with M. T. F. Wong, D. D. Harkness, R. Bol, and K. Coleman. inputs of low molecular weight carbon compounds to soil. 1999. Estimating net primary production from measurements Soil Biology & Biochemistry 42:896–903. made on soil organic-matter. Ecology 80:2762–2773. Farrar, J. E., M. Hawes, D. L. Jones, and S. Lindow. 2003. Keiser, S. D., D. A. Keiser, M. S. Strickland, and M. A. How roots control the flux of carbon to the rhizosphere. Bradford. 2014. Disentangling the mechanisms underlying Ecology 84:827–837. functional differences between decomposer communities. Fernandez, I., N. Mahieu, and G. Cadisch. 2003. Carbon Journal of Ecology 102:603–609. isotope fractionation during decomposition of plant materi- Kirchman, D., E. Knees, and R. Hodson. 1985. Leucine als of different quality. Global Biogeochemical Cycles incorporation and its potential as a measure of protein- 17:1075–1085. synthesis by bacteria in natural aquatic systems. Applied and Fierer, N., M. A. Bradford, and R. B. Jackson. 2007. Toward Environmental Microbiology 49:599–607. an ecological classification of soil bacteria. Ecology 88:1354– Lajtha, K., R. D. Bowden, and K. Nadelhoffer. 2014. Litter 1364. and root manipulations provide insights into soil organic 472 JOHANNES ROUSK AND SERITA D. FREY Ecological Monographs Vol. 85, No. 3

matter dynamics and stability. Soil Science Society of Schiff, S., R. Aravena, S. Trumbore, M. Hinton, R. Elgood, America Journal 78:S261–S269. and P. Dillon. 1997. Export of DOC from forested Le Guillou, C., D. A. Angers, P. A. Maron, P. Leterme, and S. catchments on the Precambrian shield of central Ontario: Menasseri-Aubry. 2012. Linking microbial community to soil clues from C-13 and C-14. Biogeochemistry 36:43–65. water-stable aggregation during crop residue decomposition. Schlesinger, W. H., and J. A. Andrews. 2000. Soil respiration Soil Biology & Biochemistry 50:126–133. and the global carbon cycle. Biogeochemistry 48:7–20. Moore, J. C., et al. 2004. Detritus, trophic dynamics and Schmidt, M. W. I., et al. 2011. Persistence of soil organic matter biodiversity. Ecology Letters 7:584–600. as an ecosystem property. Nature 473:49–56. Neutel, A. M., J. A. P. Heesterbeek, J. van de Koppel, G. Six, J., S. D. Frey, R. K. Thiet, and K. M. Batten. 2006. Hoenderboom, A. Vos, C. Kaldeway, F. Berendse, and P. C. Bacterial and fungal contributions to carbon sequestration in de Ruiter. 2007. Reconciling complexity with stability in agroecosystems. Soil Science Society of America Journal naturally assembling food webs. Nature 449:599–602. 70:555–569. Newell, S. Y., and R. D. Fallon. 1991. Toward a method for Sollins, P., M. G. Kramer, C. Swanston, K. Lajtha, T. Filley, measuring instantaneous fungal growth-rates in field sam- A. K. Aufdenkampe, R. Wagai, and R. D. Bowden. 2009. ples. Ecology 72:1547–1559. Sequential density fractionation across soils of contrasting Nilsson, L. O., E. Ba˚a˚th, U. Falkengren-Grerup, and H. mineralogy: evidence for both microbial- and mineral- Wallander. 2007. Growth of ectomycorrhizal mycelia and controlled soil organic matter stabilization. Biogeochemistry composition of soil microbial communities in oak forest soils 96:209–231. along a nitrogen deposition gradient. Oecologia 153:375–384. Strickland, M. S., C. Lauber, N. Fierer, and M. A. Bradford. Reischke, S., J. Rousk, and E. Ba˚a˚th. 2014. The effects of 2009. Testing the functional significance of microbial glucose loading rates in bacterial and fungal growth in soil. community composition. Ecology 90:441–451. Soil Biology & Biochemistry 70:88–95. Strickland, M. S., and J. Rousk. 2010. Considering fungal : bac- Romani, A. M., H. Fischer, C. Mille-Lindblom, and L. J. terial dominance in soils—methods, controls, and ecosystem Tranvik. 2006. Interactions of bacteria and fungi on implications. Soil Biology & Biochemistry 42:1385–1395. decomposing litter: differential extracellular enzyme activi- Sulzman, E. W., J. B. Brant, R. D. Bowden, and K. Lajtha. ties. Ecology 87:2559–2569. 2005. Contribution of aboveground litter, belowground litter, Rousk, J., and E. Ba˚a˚th. 2007. Fungal and bacterial growth in and rhizosphere respiration to total soil CO2 efflux in an old soil with plant materials of different C/N ratios. FEMS growth coniferous forest. Biogeochemistry 73:231–256. Microbiology Ecology 62:258–267. Thiet, R. K., S. D. Frey, and J. Six. 2006. Do growth field Rousk, J., and E. Ba˚a˚th. 2011. Growth of saprotrophic fungi efficiencies differ between microbial communities in and bacteria in soil. FEMS Microbiology Ecology 78:17–30. fungal : bacterial ratios? Reality check and methodological Rousk, J., P. C. Brookes, and E. Ba˚a˚th E. 2009. Contrasting issues. Soil Biology & Biochemistry 38:837–844. soil pH effects on fungal and bacterial growth suggests Waksman, S. A., and F. C. Gerretsen. 1931. Influence of functional redundancy in carbon mineralization. Applied and temperature and moisture upon the nature and extent of Environmental Microbiology 75:1589–1596. decomposition of plant residues by microorganisms. Ecology Rousk, J., P. W. Hill, and D. L. Jones. 2015. Priming of the 12:33–60. decomposition of ageing soil organic matter: concentration Waksman, S. A., and R. L. Starkey. 1932. Partial sterilization dependence and microbial control. Functional Ecology of soil, microbiological activities and soil fertility. II. Soil 29:285–296. Science 16:247–268. Rousk, K., J. Rousk, D. L. Jones, O. Zackrisson, and T. H. Wardle, D. A., R. D. Bardgett, J. N. Klironomos, H. Seta¨la¨, DeLuca. 2013. Feather moss nitrogen acquisition across W. H. van der Putten, and D. H. Wall. 2004. Ecological natural fertility gradients in boreal forests. Soil Biology & linkages between aboveground and belowground biota. Biochemistry 61:86–95. Science 304:1629–1633. Ruess, L., and P. M. Chamberlain. 2010. The fat that matters: Wieder, W. R., G. B. Bonan, and S. D. Allison. 2013. Global soil food web analysis using fatty acids and their carbon soil carbon predictions are improved by modelling microbial stable isotope signal. Soil Biology & Biochemistry 42:1898– processes. Nature Climate Change 3:909–912. 1910. Williams, M. A., D. D. Myrold, and P. J. Bottomley. 2006. Sayer, E. J., M. S. Heard, H. K. Grant, T. R. Marthews, and Carbon flow from 13C-labelled straw and root residues into E. V. J. Tanner. 2011. Soil carbon release enhanced by the phospholipid fatty acids of a soil microbial community increased tropical forest litterfall. Nature Climate Change under field conditions. Soil Biology & Biochemistry 38:759– 1:304–307. 768.