Continental-scale patterns of extracellular enzyme activity in the subsoil: an overlooked reservoir of microbial activity

Chemical stabilization of microbial-derived products such as extracellular enzymes (EE) onto mineral surfaces has gained attention as a possibly important mechanism leading to the persistence of soil organic carbon (SOC). While the controls on EE activities and their stabilization in the surface soil are reasonably well-understood, how these activities change with soil depth and possibly diverge from those at the soil surface due to distinct physical, chemical, and biotic conditions remains unclear. We assessed EE activity to a depth of 1 m (10 cm increments) in 19 soil profiles across the Critical Zone Observatory Network, which represents a wide range of climates, soil orders, and vegetation types. For all EEs, activities per mass of soil correlated positively with microbial biomass (MB) and SOC, and all three of these variables decreased logarithmically with depth (p < 0.05). Across all sites, over half of the potential EE activities per mass soil consistently occurred below 20 cm for all measured EEs. Activities per unit MB or SOC were substantially higher at depth (soils below 20 cm accounted for 80% of whole-profile EE activity), suggesting an accumulation of stabilized (i.e. mineral sorbed) EEs in subsoil horizons. The pronounced enzyme stabilization in subsurface horizons was corroborated by mixed-effects models that showed a significant, positive relationship between clay concentration and MB-normalized EE activities in the subsoil. Furthermore, the negative relationships between soil C, N, and P and C-, N-, and P-acquiring EEs found in the surface soil decoupled below 20 cm, which could have also been caused by EE stabilization. This finding suggests that EEs may not reflect soil nutrient availabilities deeper in the soil profile. Taken together, our results suggest that deeper soil horizons hold a significant reservoir of EEs, and that the controls of subsoil EEs differ from their surface soil counterparts.


Introduction
Globally, soils store approximately 1500 Pg of soil organic carbon (SOC) in the upper meter of the soil profile, with 50%-67% of SOC occurring below 20 cm (Jobbágy and Jackson 2000). The persistence of this C pool is, in part, controlled by extracellular enzymes (EEs) primarily released by soil microorganisms that decompose soil organic matter (Burns et al 2013). However, even though the majority of SOC occurs in the subsoil, most studies of soil microorganisms and the EEs they secrete focus on the upper soil layers (Yost and Hartemink 2020). While the age (and thus persistence) of SOC increases with depth (Trumbore et al 1996, Paul et al 1997, Rumpel et al 2002, recent studies have shown that subsoil (>20 cm depth) C is still vulnerable to decomposition. Indeed, subsoil microbial communities have resource demands that rival those of surface soils when normalized to a microbial biomass (MB) basis (Jones et al 2018). Understanding subsurface processes is critical in an age of global change because vulnerability of SOC to EE attack could be enhanced by increased temperatures or wetting/drying cycles (Schimel et al 2011, Hicks Pries et al 2017. Hence, if subsoils are disturbed (either physically or through altered environmental conditions), portions of the soil organic matter pool at depth could become accessible to EEs, resulting in the mineralization of significant quantities of C and nutrients. Therefore, increased understanding of EE patterns at depth could help elucidate the mechanisms of subsoil organic matter decomposition and aid in predicting how pools of SOC and nutrients will be affected by ongoing global change factors.
Because EEs both respond to and influence soil properties, the study of EEs has led to greater insights into soil C persistence (Billings and Ballantyne 2013, Birge et al 2015, Dove et al 2019, nitrogen (N) and phosphorus (P) mineralization , Waring et al 2014, ecosystem development (Olander and Vitousek 2000, Selmants and Hart 2010, Turner et al 2014, and microbial metabolism (Sinsabaugh and Shah 2011. Given that the methods for measuring EE activity in soils are relatively high-throughput, inexpensive, and reproducible across laboratories , it is one of the most common soil biogeochemical measurements ('soil extracellular enzyme activity' resulted in 2013 records using Clarivate Analytics Web of Science as of Jan. 28, 2020). However, despite the widespread measurement of soil EEs, most studies have focused on EE activities in surface horizons, with few studies exploring EE activity patterns in soil horizons below 20 cm (but see Taylor et al 2002, Kramer et al 2013, Stone et al 2014, Taş et al 2014, Schnecker et al 2015, Loeppmann et al 2016, Jing et al 2017. Numerous soil physical and biogeochemical properties change with depth. As organic matter (both SOC and organically bound nutrients) moves into the subsoil, it becomes increasingly more microbially processed and sorbed onto charged mineral surfaces (Rumpel and Kögel-Knabner 2010), which concomitantly increase with depth. Soil pH may also increase with depth in instances where the parent material is enriched in so-called 'base' cations (i.e. calcium, magnesium, potassium, and sodium ;Brubaker et al 1993). These gradients in soil properties result in subsoil microbial communities that are vastly different than their surface soil counterparts (Eilers et al 2012, Brewer et al 2019. Soil pH (Sinsabaugh et al 2008, Kivlin andTreseder 2014), substrate availability and demand (Olander andVitousek 2000, Dove et al 2019), and microbial community composition (Schnecker et al 2015) influence EE activities in surface soils. Because these factors change along soil profiles, EE activities should also change with soil depth. Two main generalizations have emerged from the few studies that have investigated EE activities in subsoils: (1) EE activities decline with depth in association with decreases in soil organic matter concentrations and decreases in MB (Taylor et al 2002, Stone et al 2014, Loeppmann et al 2016; and (2) EE activities at depth are less responsive to surface conditions, manipulations, and management practices (Kramer et al 2013, Jing et al 2017, Yao et al 2019. However, our ability to quantify the total EE pool and elucidate the controls on EEs in subsoils has been hindered by unstandardized ancillary measurements, assay parameters, and depths of sampling across studies (Nannipieri et al 2018).
Systematic, continental-and global-scale assessments and meta-analyses of EEs in surface soils have begun to clarify controls and correlates of EE activity (Sinsabaugh et al 2008, 2009. For instance, EE stoichiometry (the ratio of C-, N-, and P-acquiring enzymes), which can represent the relative C, N, and P demand (Sinsabaugh and Shah 2012), scales at 1:1:1 (C:N:P) globally across soil, freshwater, and saltwater ecosystems, suggesting that the plasticity of microbial resource demand is somewhat constrained (Sinsabaugh et al 2008(Sinsabaugh et al , 2009). These large-scale assessments also confirm that pH, substrate availability, and microbial demand influence EE activity in surface soils (Sinsabaugh et al 2008, 2009. However, it is currently unknown if these controls in surface soils extend into the subsoil. We posit that EE activities at depth may follow different patterns than in the surface horizons given that EEs at depth are less responsive to environmental perturbations (Jing et al 2017), subsoils have greater spatial heterogeneity of organic substrates than at the surface (Salomé et al 2010), and the microbial communities at depth are dominated by oligotrophic microorganisms (Brewer et al 2019).
To quantify EE activities and elucidate their controls throughout the soil profile, we sampled the upper meter of mineral soil at 10 cm increments in 19 soil pits across the 10 United States National Science Foundation-supported Critical Zone Observatories (CZOs; http://criticalzone.org/national/). We hypothesized that EE activities per mass of soil would decline with depth due to decreased SOC and MB concentrations; however, a significant proportion of EE activity in the top meter of soil would occur below 20 cm depth. We also hypothesized that the fundamental controls on EE activities would differ between surface and subsoil horizons due to shifting biological, chemical, and physical conditions throughout the soil profile. Specifically, as organically bound microbial resources decrease with depth, mineral sorption of both substrates and EEs will become a more dominant control of potential EE activity. Our overall goal was to quantify potential EE activity in the subsoil over a diverse set of soils, ecosystems, and climates to elucidate how EE activity mediates subsoil C and limiting nutrient availabilities.

Site selection and sampling
Samples were collected from the network of ten CZOs (http://criticalzone.org) across the USA, which represents a wide range of hydrogeological provinces, soil orders, and vegetation types as described in Brewer et al (2019). Soils were collected at peak greenness (as estimated from NASA's MODerate-resolution Imaging Spectroradiometer, or MODIS) between April 2016 and November 2016, with the exception of the Eel River CZO samples, which were collected in May 2017 (also at peak-greenness). At each CZO, we excavated two separate soil profiles ('sites') selected to represent distinct soil types and landscape positions (table 1). Any organic horizon was first removed, and then mineral soils were sampled in 10 cm increments with a sterile hand trowel dug into the face of each soil pit to a depth of at least 100 cm or to refusal (e.g. bedrock, hardpan, coarse regolith).
All soil samples were shipped overnight at 4 • C to the University of California, Riverside for processing. A portion of each field sample was sieved (<2 mm), homogenized, divided into subsamples for further analyses, and frozen (−20 • C). For some soils (particularly some wet, finely textured depth intervals), sieving was impractical. These samples were homogenized and larger root and rock fragments were removed by hand. In addition, as samples from SHAL (70-100 cm depth; see table 1 for site abbreviations) consisted almost entirely of medium-sized weathered bedrock (Cr material), soil was collected by manually crushing weathered bedrock with a ceramic mortar and pestle with this material then passed through a 2 mm sieve.

Soil physiochemical measurements
Soil pH, gravimetric water content, and clay concentration were measured using modified long-term ecological research protocols (Robertson et al 1999). Briefly, soil pH was determined in a 1:2 (weight to volume) solution using 5 g of oven-dried soil and 10 ml of Milli-Q water (Millipore Sigma, Burlington, MA, USA). The solution was measured on an Orion DUAL STAR pH meter and an epoxy combination electrode (Orion 9165BNWP Combination Sure-Flow pH Electrode; Thermo Fisher Scientific, Waltham, MA, USA). For determining gravimetric water content, approximately 7 g field-moist soil was dried at 105 • C for a minimum of 24 h. Soil texture was measured on oven-dried and sieved soil using the hydrometer method following Gee and Bauder (2018).
Prior to soil total organic C and N analysis, soils were freeze-dried using a Savant Novalyphe-NL500 freezer dryer (Savant, Farmingdale, NY, USA) and ground to a fine powder using a roller mill. If effervescence occurred when a drop of 1 M HCl was added to a subsample of each soil sample, then inorganic C was removed from 2 g of the soil sample by twicewashing with 30 ml 0.1 N HCl (allowing the soil slurry to stand for 1 h during each wash), twice-washing with 30 ml DI, and then freeze-dried. The soil samples were analyzed for total organic C and total N by continuous-flow, direct combustion using a Vario Micro Cube elemental analyzer (Elementar, Hanau, Germany).
Microbially available orthophosphate, referred hereafter as Olsen P, was estimated by extracting 1 g of soil with 200 ml of 0.5 M NaHCO 3 at pH 8.5 (Olsen et al 1954). This measurement includes both directly available phosphate and phosphate bound to calcium minerals that could become potentially available to microbes. Briefly, slurries were shaken for  30 min and filtered through Whatman No. 42 filters. Orthophosphate was measured colormetrically using a Lachat AE Flow Injection Auto Analyzer (Method 12-115-01-1-Q, Lachat Instruments, Inc. Milwaukee, WI, USA).

Phospholipid fatty acid analysis
We used phospholipid fatty acids (PLFAs) to determine differences in the MB and the ratios of fungal to bacterial biomass. Briefly, total lipids were extracted using 10 ml of methanol, 5 ml chloroform, and 4 ml of a 50 mM phosphate buffer (pH = 7.4) from 5 g of lyophilized soil (White et al 1979, Deforest et al 2004. To determine analytical recovery, phospholipid 19:0 (1,2-dinonadecanoyl-sn-glycero-3phosphocholine) and 21:0 (1,2-diheneicosanoyl-snglycero-3-phosphocholine) standards (Avanti Polar Lipids, Inc. Alabaster, AL, USA) were added during the extraction phase (Deforest et al 2012). Polar lipids were separated from other lipids using silicic acid solid-phase chromatography columns (500 mg 6 ml −1 ; Thermo Scientific, Waltham, MA, USA), and the separated polar lipids were converted to fatty acid methyl esters (FAME) through methanolysis (Guckert et al 1985). The resulting FAMEs were separated using a HP GC-FID (HP6890 series, Agilent Technologies, Inc. Santa Clara, CA, USA) gas chromatograph, and peaks/biomarkers were identified using the Sherlock System (v. 6.1, MIDI, Inc. Newark, DE, USA). External FAME standards (K104 FAME mix, Grace, Deerfield, IL, USA) were used to determine concentrations. The sum of all detected 14-19 C-length PLFAs was used to calculate MB because longer PLFAs can be indicators of mosses and higher plants (Zelles 1999). Ratios of fungal to bacterial biomass (fungi:bacteria) were calculated by dividing the amount (mol) of the fungal biomarker 18:2ω6c by the sum of all other microbial biomarkers (i.e. mol 18:2ω6c/(mol MB-mol 18:2ω6 c)).
The choice of buffer pH in EE activity assays depends on the research question (Burns et al 2013) and, as such, we decided to use a consistent pH of 5.5 for all soils assayed similar to other crosssite soil EE studies (Deforest 2009. The intensity of florescence of MUB is pH dependent (Mead et al 1955), therefore comparisons across sites must be done at a consistent pH to avoid attributing biological phenomena to the chemistry of the florescent substrate. A buffer pH of 5.5 was chosen because this is within the range of soil pH for most of our sites (table 1) and the range of pH optima for our enzymes (4.0-6.5, as determined in a variety of biomes; Parham and Deng 2000, Niemi and Vepsäläinen 2005, Turner 2010, Min et al 2014. Nevertheless, we recognize that our buffer pH may not be indicative of the native soil pH (or pH within the microsites in which EEs operate), which may reduce our ability to quantify in situ EE activity.
The enzymes, AG, BG, BX, and CB are involved in the degradation of organic C, and total C-acquiring enzyme activity (C sum ) was operationally defined as the sum of these four enzyme activities. The enzyme, NAG is involved in releasing N-acetylglucosamine from aminopolysaccharides such as chitin and peptidoglycan, and LAP catalyzes the hydrolysis of leucine residues at the N-terminus of peptides and proteins. Both NAG and LAP are considered N-acquiring enzymes and were similarly summed to define the variable N sum , which we use as a proxy for N acquisition by decomposition. Acid phosphatase is involved in releasing phosphate from monoester bonds, representing a P-mineralizing enzyme (Burns et al 2013). This suite of EEs, while not inclusive of all relevant enzymatic substrates, represents many of the most frequent hydrolytic reactions during decomposition of organic matter (Sinsabaugh and Shah 2012). Furthermore, these EEs have been extensively studied across numerous surface soils (Sinsabaugh et al 2008(Sinsabaugh et al , 2009) to which we can compare with our deep-soil measurements.
Extracellular enzyme activities were expressed per soil mass (mmol EE activity kg −1 soil h −1 ), SOC (mmol EE activity kg −1 SOC h −1 ), and MB (mmol EE activity kg −1 MB h −1 ). These latter two variables are called SOC-normalized and MB-normalized, respectively, in this paper. We also measured the ratio of C-, N-, and P-acquiring enzymes. Because EEs mediate nutrient acquisition for soil microorganisms, they can be used to determine relative nutrient demand (Olander and Vitousek 2000, Sinsabaugh and Shah 2012). Hence, we used C sum :N sum , C sum :AP, and N sum :AP as proxies for C:N, C:P, and N:P relative demand ratios, respectively.

Statistical analysis
All statistical tests and visualizations were conducted in R (R Development Core Team 2008) using the lme4 (Bates et al 2015) and MuMin (Barton 2020) packages. We used mixed-effects models with site as a random effect to examine the relationship between depth, SOC, MB, clay, and fungi:bacteria and EE activity (expressed on soil mass, SOC, and MB bases). We similarly used mixed-effects models with site as a random effect to examine the effect of soil stoichiometry (using ratios of SOC, total N, and available P) on enzyme stoichiometry. These models were conducted on the complete dataset, the surface soil dataset (depth ≤ 20 cm), and the subsoil dataset (depth > 20 cm) to determine differences in the controls of EE activities between the surface and subsoils. Because we did not characterize the horizonation of the sampling pits, we a priori chose 20 cm to represent the subsoil because most EE studies do not sample below this depth. However, we also conducted our analysis using a 30 cm threshold, and the overall interpretation remained unchanged (see appendix A). Therefore, for clarity, we report results using only the 20 cm threshold for the subsoil. To denote the variance explained by the models, we report the marginal R 2 value, which expresses the increase in explained variance by including the fixed effect(s) (Nakagawa and Schielzeth 2013). We also used ANOVA and Pearson's correlation to determine if the fraction of EE activity below 20 cm differed by soil order and if the aggregate surface and subsoil EE activities were correlated, respectively. We assessed significance at the α = 0.05 level and marginal significance at α = 0.10. If significant differences were detected, we used Tukey's Test of Honest Significant Differences to determine which soil orders were significantly different.
We used QQ-plots and scale-location plots to inspect normality and homoscedasticity, respectively. Because many of the mixed-effects models failed to meet parametric assumptions, all dependent and independent continuous variables were natural logtransformed and re-analyzed. The resulting models, along with the ANOVAs, met the assumptions of parametric tests. For visualization purposes, data are left untransformed unless otherwise stated.

Whole profile soil properties among sites
Soil organic C, total N, available P, and fungi:bacteria decreased while clay percentage increased with depth across the CZO network (all: p < 0.001, figures S1(A)-(E) (available online at stacks.iop.org/ERL/15/1040a1/mmedia)). Across all sites, soil pH increased slightly with depth (on average 0.1 pH units over 1 m, p = 0.028, figure S1(F)).

Distribution of EE activity is related to MB and organic carbon throughout the top meter of soil
For all assayed EEs, EE activity per mass of soil declined logarithmically with depth (p < 0.001, figures 1 and S2), with the strongest decline for NAG (β = −0.223) and the weakest for AG (β = −0.109). However, about 50% of the total-profile EE activity kg −1 soil in the top meter occurred below 20 cm (figure 2(A)). The proportion of the EE activity below 20 cm differed by the soil order for many of the assayed EEs (tables S1 and S2). Mollisols had about a 1.5 times greater percentage of the sum of C-and N-acquiring EE activity kg −1 soil below 20 cm than Inceptisols or Ultisols (p < 0.05 for all comparisons, figure 2(B)). For AP, the 39% higher proportion in the subsoil for Mollisols compared to Inceptisols was only marginally significant (p = 0.063). Neither mean annual temperature (MAT) nor precipitation (MAP) significantly correlated with the proportion of EE activity below 20 cm (p > 0.05, figures S3 and S4).
There were also differences in the percentage of MB and SOC in the subsoil among soil orders (MB: p < 0.001, SOC: p = 0.013), with Mollisols having an almost two times greater proportion of MB and SOC below 20 cm than Inceptisols (MB: p = 0.006, SOC: p = 0.013; figure 2(B)). While the proportion of MB below 20 cm was significantly higher in Mollisols compared to Ultisols (about 1.5 times greater, p = 0.001), the difference in the proportion of SOC below 20 cm between Mollisols and Ultisols was only marginally significant (p = 0.057).
Microbial biomass-normalized EE activity increased with depth for all enzymes (figure S5; all: p < 0.05). The strongest increases were for LAP and AP, which increased six-and seven-fold, respectively, from the 0-10 cm to the 90-100 cm depth, while NAG and BG increased by 85% and 103%, respectively. Throughout the top meter, over 80% of MBnormalized EE activity occurred below 20 cm (figure S6(A)). However, because the proportion of MB below 20 cm also varied among soil orders, the proportion of MB-normalized EE activity below 20 cm was consistent among soil orders for most assayed EEs (AG: p = 0.333, BG: p = 0.175, CB: p = 0.278, BX: p = 0.211, NAG: p = 0.027, LAP: p = 0.537, AP: p = 0.048; figure S6(A)). Nevertheless, the proportion of MB-normalized NAG activity below 20 cm was 15% greater in Ultisols compared to Inceptisols (p = 0.025), and the proportion of MB-normalized AP activity below 20 cm was 17% greater in Ultisols compared to Mollisols (p = 0.042).
There were inconsistent patterns of EE activity normalized by SOC with depth. Nacetylglucosaminidase normalized by SOC decreased with depth (p = 0.004); AG, LAP, and AP increased with depth (AG: p = 0.016, LAP: p = 0.002, AP: p < 0.001); and BG, CB, and BX did not change with depth (BG: p = 0.322, CB: p = 0.344, BX: p = 0.198; figure S7). Similar to the proportion of MB-normalized EE activity below 20 cm, the proportion of EE activity normalized by SOC below 20 cm averaged about 80% and did not differ among soil orders (all: p > 0.1; figure S8).
With a few exceptions, aggregate EE activity (per mass of soil, MB, and SOC) below 20 cm correlated with the aggregated activity in the upper 20 cm (table  S3). On average, these correlations were strongest for SOC-normalized EE activities and weakest for MBnormalized EE activities. As such, aggregate surface soil AG, CB, and BX activity normalized by MB was not correlated with respective aggregate activities in the subsoil (p > 0.05).

Controls on EE activity throughout the top meter of soil
Consistently, MB, SOC, and fungi:bacteria were better predictors of EE activities per mass of soil than pH or clay concentrations (table S4). This was generally consistent among surface soil-and subsoil-only datasets except for fungi:bacteria, which was only a strong predictor in the surface soil (table S5).
Normalized by MB, soil pH was generally not a significant predictor of the assayed EE activities (table  S6). This pattern was mostly consistent among surface soil-and subsoil-only datasets, with the exception of surface soil CB (p = 0.023) and subsoil LAP (p = 0.042, table 2). In contrast, normalized by SOC, soil pH had a variable effect on EE activities. In the surface soil, pH was positively correlated with BG (p = 0.001), CB (p = 0.002), BX (p = 0.042), and LAP (p = 0.004, table 2). However, in the subsoil, pH was negatively correlated with CB (p = 0.025) and AP (p < 0.001), and positively correlated with LAP (p < 0.001, table 2).
When EE activities were normalized per unit MB, clay concentrations and fungi:bacteria were generally correlated positively with EE activities (table  S6). When surface and subsoil EE data were analyzed separately, the effect of clay concentrations and fungi:bacteria on MB-normalized EE activities was more often significant in the subsoil (table 2).

Relating soil and EE stoichiometries throughout the top meter of soil
When considering soils from all depth increments, only soil C:N and EE C:N were correlated (C:N: p = 0.013, C:P: p = 0.292, N:P: p = 0.276), but this negative correlation between soil C:N and EE C:N was relatively weak (marginal R 2 = 0.038; figure  S9). However, using the surface soil-only dataset, all soil and EE stoichiometries were negatively correlated (C:N: p = 0.003, marginal R 2 = 0.268; C:P: p = 0.002, marginal R 2 = 0.193; N:P: p = 0.004, marginal R 2 = 0.260; figure 3). In the subsoil, these correlations decoupled such that none of the stoichiometries were significantly correlated (C:N: p = 0.288, C:P: p = 0.358, N:P: p = 0.282; figure 3). Split amongst 10 cm increment sample depths, negative correlations between soil and enzyme stoichiometry were generally significant (p < 0.05) only in the upper soil layers (figure S10).

Discussion
Our continental-scale sampling efforts show that microbial activity at depth is non-negligible, and the relative proportion of EE activity (kg −1 soil) at depth depends predominately on soil development (i.e. soil order; figure 2(B)). Although replication of each soil order was relatively small (n = 4-5), this finding was strikingly consistent despite large gradients in MAT and MAP for each soil order (e.g. MAP spanned an order of magnitude for Inceptisols; table 1). Our analysis shows that climate is an unlikely driver of the relative vertical distribution of EE activity. Instead, this phenomenon is likely due to changes in the vertical distribution of substrate (organic C) and MB among these soil orders (Batjes 1996; figure 2(B)), which strongly correlate with EE activity (Sinsabaugh et al 2008;table S4). Hence, we show that SOC and MB are the strongest controls of EE activities throughout the soil profile.
We hypothesize that increases in the MBnormalized EE activities at depth suggest an accumulation of EEs chemically stabilized on mineral and organic surfaces. While MB-normalized EE activity is often related to the relative activity of the microbial community or differences in metabolic strategies among microbial taxa (Boerner et al 2005), we alternatively hypothesize that the increase in MBnormalized EE activity is due to EE stabilization, namely the sorption of the EEs onto clay or organic matter particles that impedes EE degradation (Sarkar et al 1989, Burns et al 2013. Because EE activities are often measured in a salt-buffered soil slurry that disrupts the stabilization of EEs (as is the case in our study), EE activity assays generally measure both active and stabilized EEs (Burns et al 2013). We hypothesize that higher subsoil MB-normalized EE activities is primarily a product of EE stabilization instead of differences in the metabolic capabilities Table 2. Marginal R 2 values for mixed-effects models with soil clay concentration, pH, or fungi:bacteria ratio as the sole fixed effect, and soil pit as a random effect on extracellular enzyme (EE) activity normalized by microbial biomass (MB) or soil organic carbon (SOC) concentration in surface-(≤20 cm) and sub-soils (>20 cm) across all sites. Key: α-glucosidase (AG), β-glucosidase (BG), cellobiohydrolase (CB), β-xylosidase (BX), N-acetylglucosamine (NAG), leucine aminopeptidase (LAP), and acid phosphatase (AP) activity. Bolded values represent a significant (α = 0.05) effect and ± signifies the direction of the effect (surface soil: n = 29, subsoil: n = 114).  Secondly, the relative abundance of fungi, which produce more EEs per unit MB than bacteria (Romaní et al 2006), decreased with depth. Finally, the decoupling of soil stoichiometry and EE stoichiometry at depth suggests that EE activities are not responsive to altered nutrient availabilities. Taken together, these results suggest that the physiochemical process of EE stabilization, a largely abiotic process, is the major control of EE activity in the subsoil. Extracellular enzyme stabilization as a major mechanism in the subsoil is corroborated by our finding that the influence of clay concentration on MBnormalized EE activity is higher in the subsoil than the surface soil (table 2). Furthermore, we may have underestimated EE activity in high clay soils because clay can increase the pH optima of EEs 1-2 pH units Estermann 1957, Ramírez-Martínez andMclaren 1966). Whereas many EEs have native pH optima between 4.0 and 6.5 (Parham and Deng 2000, Niemi and Vepsäläinen 2005, Turner 2010, Min et al 2014, an increase of two pH units would be significantly higher than the pH of our assay buffer (pH = 5.5). Therefore, we conclude that EE stabilization is a major process when microbial activity is relatively low and clay concentrations are relatively high, which is often the case in subsurface soil layers.

MB-normalized
Extracellular enzyme stabilization may be partially responsible for the muted treatment effects on subsoil EE activity commonly found throughout the literature (e.g. Kramer et al 2013, Jing et al 2017, Yao et al 2019. When the stabilized EE pool is significantly greater than the active EE pool, the ability to detect changes in the active pool is decreased. For example, if we assume that there is negligible EE stabilization in the surface soil and that the actualized MB-normalized EE activity in situ is constant throughout the soil profile, our results show that at least 29%-71% of the assayed MB-normalized EE activity at depth can be attributed to stabilized EEs across our study sites, depending on the EE (equation (1)).
(1) X = Average MB-normalized EE activity in surface soil Y = Average MB-normalized EE activity in subsoil Z = Percent MB-normalized EE activity in subsoil attributed to stabilized EEs This equation calculates the difference between MB-normalized EE activity in surface and subsoil as a percentage of the MB-normalized EE activity in the subsoil and, adhering to the aforementioned assumptions, represents the percentage of MB-normalized EE activity in the subsoil attributed to EE stabilization. This calculation likely represents the lower bound of the estimated stabilized MB-normalized EE activity because any stabilization in the surface soil and nutrients) and clay concentrations. The differences in the size of boxes between the surface and subsoil represent the relative size of the pool, and differences in the thickness of arrows between the surface and subsoil represent the hypothesized relative magnitude of the flux or control. A portion of the substrate pool is available to microbial biomass (MB) and is moderated by clay concentration and active EEA. Substrate availability moderates substrate demand. Bacterial biomass, fungal biomass, and substrate demand influence active EEA. Additionally, our conceptual model incorporates stabilized EEA (i.e. EEs sorbed onto clay particles), which is primarily influenced by clay concentrations. At depth, the impact of clay on substrate availability and stabilized EEA increases, while the absolute impact of substrates and microbial properties (i.e. microbial biomass and substrate demand) decreases.
(X), would increase Z, and the relative proportion of fungal biomass, which release comparatively more EEs than bacteria per unit MB (Romaní et al 2006), decreased with depth. Nevertheless, this implies that if the stabilized EE pool is resistant to treatment effects in experiments (e.g. Kramer et al 2013, Jing et al 2017, Yao et al 2019, the ability to detect significant changes in microbial activity at depth using EE assays is also reduced by at least 29%-71%. In instances where the magnitude of the treatment effect is modest, it is unlikely that a significant change in subsoil EE activity will be detected. However, this should not necessarily be interpreted as a lack of microbial response, and caution should be exercised in interpreting the effect of a surface manipulation or treatment on subsoil EE activity. The discrepancy between soil and EE stoichiometry at depth may also be caused by the increased discontinuity of substrates in the subsoil and the reduced ability of the microbial community to respond to changes in resource availability (Allison et al 2007). This would prevent subsoil microorganisms altering their EE stoichiometry to different nutrient conditions. Resource availability is typically higher in surface soils than in subsoils (Salomé et al 2010). Recent work in soil enzymography show that C-degrading EE activities are enriched only 0.5-2.0 mm from C-rich rhizodeposits (Ma et al 2018). The EE assays that we and most other researchers use disrupt the spatial arrangement of EEs and substrates such that our results express bulk EE activities and bulk resource concentrations, which may not be representative of more localized heterogeneity in resources.
Our finding that aggregated surface soil EE activity (normalized by mass of soil, MB, or SOC) generally correlates with aggregated activity in the subsoil suggests that it may be possible to extrapolate EE measurements at the surface into deeper layers. Interestingly, MB and SOC, which we demonstrate correlate with EE activity (per mass of soil), did not follow these same patterns. It is possible that high concentrations of EEs in the surface soils percolated into the subsoil. However, correlation of surface and deeper EE activities, instead, could be due to similarities in microbial community composition throughout the soil profile. Indeed, microbial community composition assessed by 16S rRNA gene sequencing of these same soils showed a stronger effect of soil location than soil depth (Brewer et al 2019). This finding provides further evidence of the linkages between microbial community composition and metabolic strategies in soils (e.g. Schnecker et al 2015) and demonstrates association between surface soils and subsoils.
Discrepancies in the effect of soil pH between MB-normalized and SOC-normalized EE activities likely reflects the impact of soil pH on SOC stabilization and how well our bulk MB and SOC measurements correlate with microbial-available SOC. In numerous ecosystems, low soil pH is associated with greater SOC stabilization due to an increased charge of clay minerals and Al-and Fe-oxyhydroxides resulting in an increase of their sorption capacity (Rasmussen et al 2018). Therefore, increasing pH (and decreasing SOC stabilization) likely reflects greater available SOC as a fraction of total SOC, which would result in higher SOC-normalized EE activities. Soil pH was not a significant mediator of MB-normalized activities possibly because our MB measurements better reflect the available SOC pool, given that microbial growth is generally substratelimited (Jones et al 2018). These results highlight the interactions between SOC, MB, and soil stabilization of microbial substrates and products and provide further evidence for the strong effect of soil stabilization in regulating EE activities throughout the soil profile.
Taken together, our results suggest that the relative importance of the different controls on EE activities change with depth. We summarize this in a conceptual model, where the active EE pool is controlled by microbial EE production (proximately influenced by MB and resource demand), and the stabilized EE pool is primarily influenced by EE stabilization onto clay particles (figure 4). Because MB and resource demand decrease with depth as C becomes more limiting and clay concentrations increase, the subsoil total EE pool is maintained because of the relatively large proportion of stabilized (sorbed on soil colloids) EEs that decay slower than unstabilized (present in the bulk soil solution) EEs. Understanding how soil texture affects EE stabilization and decay dynamics is a critical knowledge gap in enzyme-explicit microbial models (e.g. Schimel and Weintraub 2003, Manzoni et al 2016, Abramoff et al 2017, Sulman et al 2018. For instance, Schimel et al (2017) estimated EE decay dynamics in multiple soils by measuring EE activities for weeks after sterilization. While these soils varied in texture, there did not appear to be a consistent pattern between soil texture and EE decay, possibly because of changes in other edaphic factors (i.e. moisture, substrate, etc.). Future work should systematically study EE decay and its relation to multiple edaphic factors including clay concentration to test our proposed conceptual model.
Overall, the suite of EEs studied here exhibit similar patterns with depth across a wide range of sites and represent a diverse set of biochemical reactions. Hence, we posit that these patterns are robust and may be applicable to other EEs released by soil microorganisms. Our findings imply that the vast majority of EE studies are missing a large portion of the total EE activity in soils, and that the unmeasured subsoil EE activity varies in its response to environmental conditions. Nevertheless, if undisturbed, extrapolating surface soil EE values into the subsoil may be appropriate. As numerous other experiments have shown (Blume et al 2002, Taş et al 2014, ignoring subsoils, and exclusively focusing on surface soils, can limit our ability to understand whole-profile EEdynamics and soil C storage.

Distribution of EE activity is related to microbial biomass and organic carbon throughout the top meter of soil
About 40% of the total-profile EE activity kg −1 soil in the top meter occurred below 30 cm ( figure A1(A)). The proportion of the EE activity below 30 cm differed by the soil order for many of the assayed EEs (tables A1 and A2). Mollisols had about a 1.5 times greater percentage of the sum of carbon (C)-and nitrogen (N)-acquiring EE activity kg −1 soil below 30 cm than Inceptisols (p < 0.050 for all comparisons, figure A1(B)). For AP, differences among soil orders was only marginally significant (p = 0.057).
Neither mean annual temperature nor precipitation significantly correlated with the proportion of EE activity below 30 cm (p > 0.05, figures A2 and A3). There were also differences in the percentage of microbial biomass (MB) and soil organic C (SOC) in the subsoil among soil orders (MB: p < 0.001, SOC: p = 0.009), with Mollisols having an almost two times greater proportion of MB and SOC below 30 cm than Inceptisols (MB: p = 0.003, SOC: p = 0.008; figure A1(B)). While the proportion of MB below 30 cm was significantly higher in Mollisols compared to Ultisols (about 1.5 times greater, p < 0.001), the difference in the proportion of SOC below 30 cm between Mollisols and Ultisols was only marginally significant (p = 0.059).
Throughout the top meter, about 75% of MBnormalized EE activity occurred below 30 cm ( figure A4(A)). However, because the proportion of MB below 30 cm also varied among soil orders, the proportion of MB-normalized EE activity below 30 cm was consistent among soil orders for all assayed EEs (AG: p = 0.251, BG: p = 0.334, CB: p = 0.332, BX: p = 0.367, NAG: p = 0.081, LAP: p = 0.670, AP: p = 0.154; figure A4(A)). Similar to the proportion of MB-normalized EE activity below 30 cm, the proportion of EE normalized by SOC below 30 cm averaged about 70% and did not differ among soil orders (all: p > 0.1; figure A5).
With a few exceptions, aggregate EE activity (per mass of soil, MB, and SOC) below 30 cm correlated with the aggregated activity in the upper 30 cm (table A3). On average, these correlations were strongest for SOC-normalized EE activities and weakest for MB-normalized EE actives. As such, aggregate surface soil AG, BG, and CB, activity normalized by MB was not correlated with respective aggregate activities in the subsoil (p > 0.05).

Controls on EE activity throughout the top meter of soil
While EE activities (kg −1 soil) were always correlated with MB and SOC for both the surface and subsoil datasets, clay concentration and the ratio of fungito-bacteria (fungi:bacteria) were only correlated with EE activities in the surface soil (table A4). With the exception of surface-and subsoil AP, which positively correlated with pH, and subsoil LAP, which negatively correlated with soil pH, pH was not a significant predictor of EE activities (table A4). Normalized by SOC or MB, soil pH had a variable effect on EE activities (table A5). In the surface soil, pH correlated positively with BG and CB regardless of the normalization, while pH positively correlated with LAP only when normalized by SOC. In the subsoil soil pH positively correlated with LAP and negatively correlated with AP only when normalized by SOC.

Figure A6.
Correlations between soil and extracellular enzyme (EE) stoichiometry (i.e. the ratio of elements by mass and extracellular enzyme activities that target these same elements) of carbon (C), nitrogen (N), and phosphorus (P) in surface (≤30 cm depth) and subsoils (>30 cm depth across all sites). Blue lines show significant (α = 0.05) mixed-effects models of the relationship between soil and EE stoichiometry (site was used as a random effect; lines were not drawn where correlations were not significant). Gray ribbons show the standard error of the model. Data points represent individual soil samples (depths within each pit). Note the scales of both axes differ among plots.