Three different soils-Ultisol, Alfisol, and Vertisol-were collected from mixed forests located in the southern United States. The Ultisol belonged to the Cowart series (fine-loamy, kaolinitic, thermic Typic Kanhapludults) and was collected from Taylor County in Georgia (32.54°N, 84.22°W). The predominant tree species comprising the forest are Quercus alba, Acer saccharum, Quercus velutina, Juniperus virginiana, Carya spp., Pinus clausa, Nyssa sylvatica, Magnolia grandiflora, and Liriodendron tulipifera. Mean annual temperature and precipitation at this site is 20°C and 1,200 mm, respectively. The Alfisol belonged to the Weller series (fine, smectitic, mesic Aquertic Chromic Hapludalfs) and was collected from the Missouri Ozark AmeriFlux (MOFLUX) site in central Missouri. This site is located at the University of Missouri Baskett Wildlife Research and Education Center (38.74°N, 92.20°W) in an upland oak-hickory forest with major tree species consisting of Quercus alba, Carya ovata, Acer saccharum and Juniperus virginiana (Gu et al., 2015; Wood et al., 2018). Mean annual temperature and precipitation at this site is 12°C and 986 mm, respectively. The Vertisol belonged to the Buxin series (very-fine, smectitic, thermic Aquic Hapluderts) and was collected from Bowie County, Texas (33.44°N, 94.48°W). At this site the predominant tree species are Pinus taeda, Pinus echinata, Quercus spp, Carya spp, and Cupressus spp. Mean annual temperature and precipitation at this site is 17°C and 900 mm, respectively. Basic properties of these three soils are reported in Table 1.

Soil cores of 5 cm diameter were collected from several random locations at each site to a depth of 15 cm, after removing the litter layer, and combined to obtain a composite sample per site (Mavi et al., 2012; Jagadamma et al., 2014). Three additional undisturbed cores per site were collected to determine WHC. The soil samples were transported to the research laboratory in coolers with dry ice. The composite samples were immediately sieved through a 2 mm sieve. One section of the sieved samples was air-dried for the initial soil characterization (n = 3), and the remainder was stored at 4°C for 48 h to use for the microcosm experiment. Soil texture was determined by the Bouyoucos hydrometer method (Gee and Or, 2002). Soil pH was measured in a 1:2 soil:water suspension (Thomas, 1996) using a pH meter (Mettler Toledo, Columbus, Ohio). Total C was determined on finely ground samples by the dry combustion method using a Vario TOC cube CN analyzer in solid mode (Elementar, Hanau, Germany). As all the soils were acidic (pH < 6), total C was assumed to be equal to SOC (Al-Kaisi et al., 2005). Extractable organic C (EOC) was determined based on the method by Jones and Willett (2006). Briefly, soils were mixed with 0.5 M K₂SO₄ (1:4 soil:solution), shaken at 200 revolutions per minute for 1 h, and centrifuged at 3,500 rpm for 3 min. The supernatant was passed through a 0.45 μm filter paper (Whatman No. 42) and analyzed using a Vario TOC cube CN analyzer in liquid mode (Elementar, Hanau, Germany).

This study included five soil moisture treatments: air-dried, 25, 50, 100, and 175% WHC. The WHC of the three soils was determined using pressure plate extractor method (Klute and Dirksen, 1986) by placing the saturated soil cores in the pressure plate extractors at −0.33 bar and then measuring the gravimetric soil moisture content of the core (Mavi et al., 2012; Zhou et al., 2014; Bao et al., 2016). We also determined the WHC using the funnel method (Jenkinson and Powlson, 1976). As the WHC values obtained from both methods were similar, the WHC obtained using pressure plate extractors is reported. Except for the air-dried treatments, the moisture treatment manipulations involved adding MilliQ water to the soils based on the moisture treatments (Zhou et al., 2014). Prior to the start of incubation, soil samples were pre-incubated for 1 week at environmental conditions similar to the experiment (i.e., at room temperature (20°C) in the dark). Thirty grams of soil sample (on a dry weight basis) was added to plastic cups and MilliQ water was added to manipulate corresponding soil moisture treatments followed by gently mixing with a spatula. The cups were then placed in 1 L glass jars and closed with lids fitted with sampling ports at the center. A total of 12 replicates per treatments were prepared for each soil type to destructively harvest three replicates at days 1, 15, 60, and 90 of the incubation. Soil collected from the destructive harvests were frozen at −20°C pending analysis. To maintain constant soil moisture during the course of the incubation, water loss was checked on a weekly basis by weighing each soil container and corrected by adding MilliQ water as necessary. At the most, 0.1–0.2 ml of water was added every week.

Gas samples (15 ml) from the headspace of the jars (n = 3) were collected through the sampling port on the lids on days 0, 1, 2, 5, 7, 12, 19, 26, 34, 41, 48, 55, 62, 69, 76, 83, and 90 using a syringe-needle assembly and transferred to evacuated exetainer vials (Labco, United Kingdom). The jars were sampled before opening the lids and after gas sample collection, the jars were opened, and electrical fans were used for blowing air into the jars to maintain an aerobic incubation environment. For blank correction, gas samples were collected from triplicate empty (without soil) jars at all time points. The CO₂ concentration in the gas samples was measured using a flame ionization detector on a gas chromatograph (Shimadzu GC-2014, Japan). Microbial respiration rate was calculated using the linear response curve of CO₂ concentrations versus time (Curiel-Yuste et al., 2007).

The activities of four C-acquiring hydrolytic enzymes, α-glucosidase (AG), β-glucosidase (BG), cellobiohydrolase (CBH), and β-xylosidase (XYL), were determined from all treatments at all the destructive sampling time points (day 1, 15, 60, and 90) in 96-well plates according to German et al. (2011b). Each column of 12 wells on each plate corresponded to one soil sample. Another plate was used to create a standard curve for each sample at 25°C. Each column in the standard curve plate contained a soil slurry with a different concentration of the 4-methylumbelliferone (MUB) standard (0, 2.5, 5, 10, 25, 50, and 100 μM) in each of the wells. The reference standards for the enzymes AG, BG, CBH, and XYL are 4-MUB-α-D-glucopyranoside, 4-MUB-β-D-glucopyranoside, 4-MUB-β-D-cellobioside, and 4-MUB-β-D- xylopyranoside, respectively. A 2.75 g of frozen soil sample from each treatment was thawed to room temperature and homogenized with 91 ml of 50 mM sodium acetate buffer (pH adjusted according to the pH of the soil types shown in Table 1) for 1 min in a Waring laboratory grade blender on high speed to prepare a slurry. Homogenization was followed by adding 800 μL of the soil slurry into each of the eight wells of one column. After adding 12 samples in a plate, 200 μL of 200 μM respective substrates were added. The plates were incubated for 3 h at 25°C. After incubation, the plates were centrifuged at 350 × g for 3 min and 250 μL of the supernatant from each well was transferred to a 96-well black plate. A plate reader (Synergy-BioTek Instruments, Inc, Vermont, United States) was used to measure fluorescence at wavelengths 365 and 450 nm for excitation and emission, respectively. The standard curve plates were used to construct a linear standard curve to determine each enzymes’ activity for each sample as nmol g⁻¹ dry soil h⁻¹. Total activity of the C acquiring hydrolytic enzymes was defined as the sum of CBH, AG, BG, and XYL.

Microbial community composition was determined using the phospholipid fatty acid (PLFA) and neutral fatty acid (NLFA) analyses (Buyer and Sasser, 2012). Briefly, 1.5–2 g of freeze-dried soil were used for Bligh-Dyer lipid extraction (Bligh and Dyer, 1959). The extracted fatty acids were dried and analyzed using a gas chromatograph (Agilent 7890A, Agilent Technologies, Wilmington, DE, United States ) after trans-esterification for quantitative analysis relative to an internal standard. Fatty acid profiles were identified using MIDI PLFAD1 calibration mix and software SHERLOCK version 6.2 (MIDI, Inc, DE, United States ). Bacteria, fungi, and arbuscular mycorrhizal fungi (AMF) were targeted using the fatty acid markers, where gram-positive bacteria markers were 15:0, i15:0, a15:0, i16:0, 16:0ω9, i17:0, and cy17:0, and gram-negative bacteria markers were 16:1ω7, 18:1ω7, and cy19:0 (Frostegård and Baath, 1996). The amount of PLFA 18:2ω6 was used as a marker of non-mycorrhizal fungal abundance and NLFA 16:1ω5 for AMF (Olsson et al., 1995; Olsson, 1999).

The CO₂ data corresponding to each moisture treatment in each soil was best fitted using a double pool exponential decay model based on the best fit model parameters among single, double, and triple-pool models (lowest Akaike Information Criterion values and dependencies, and highest R²) (Farrar et al., 2012; Jagadamma et al., 2014; Das et al., 2019). C t = C 1 ( 1 − e − k 1 t ) + C 2 ( 1 − e − k 2 t ) where C_t is the total CO₂-C production per unit soil weight (μg C g⁻¹ dry soil), t is the time in days, C₁ is the fast mineralizing SOC pool (active pool), C₂ is the slow mineralizing SOC pool (slow pool) within time t, and k₁ and k₂ are the mineralization rates for C₁ and C₂, respectively. For each set of data, the model was fit using Sigma plot v14 (Systat Software Inc, IL, United States). Parameter constraints while fitting the model were k₁ > 0 and k₂ > 0 and C₁ + C₂ = 100% (Das et al., 2019).

Statistical analyses were conducted using SAS software v 9.4 (SAS Institute Inc., 2002). The effect of moisture on CO₂ and EOC over the incubation period for each soil was analyzed using repeated measures ANOVA with incubation length as the repeated measure. Post hoc comparisons for determining the effect of soil moisture on respiration was performed using PROC GLIMMIX in SAS. The mean separation was done using Tukey’s test. The dependence of observed microbial respiration to various soil properties including soil pore space, bulk density, SOC, pH, MBC, EOC and contents of sand, silt and clay was determined using multiple linear regression analysis by applying stepwise model selection procedure. In all statistical tests, the mean differences were considered significant if p ≤ 0.05. Moisture sensitivity for enzyme assays was estimated using regression analyses of measured enzyme activities for the moisture treatments, soil types, each destructive sampling time, and the total incubation length. To compare enzyme activities among soil moisture treatments on each day of destructive sampling, one-way ANOVA was used.

To understand how the microbial community composition varies across various moisture levels in all soil types, permutational multivariate analysis of variance (PERMANOVA) was used in R software (R Development Core Team, 2011) on the matrix of abundances of each microbial type for all soil types and moisture treatments. The Bray-Curtis distance metric and 9,999 permutations of residuals for the model were used (Kivlin et al., 2019). The Mantel test was also performed in R to assess correlations between enzyme activity and microbial community composition.

The CO₂ respiration rate showed significant responses to soil moisture content in all soils (Figure 1) (p < 0.0001–0.032 for Ultisol; < 0.0001–0.004 for Alfisol; < 0.0001 for Vertisol across the time points). The lowest respiration rate was observed at the lowest moisture content (air-dried soils) in all the three soils throughout the incubation. However, the highest CO₂ respiration rate was observed at different moisture levels for the different soil types. For Ultisol, after the first 5 days of incubation, the highest respiration rate was consistently observed at 50% WHC throughout the incubation (Figure 1A). For Alfisol, after a week of incubation, the highest respiration was observed at 100% WHC for the remaining incubation period (Figure 1B). For Vertisol, the highest respiration rate was observed at 175% WHC consistently throughout the incubation period (Figure 1C). Also, we observed a slight increase in respiration rate around day 34 in all soils, however, the trendline over time did not show an increase in any of the soils. The repeated measures ANOVA results showed significant effects of soil moisture (p < 0.0001), time of sampling (p < 0.0001), and their interaction (p < 0.001) on soil respiration for all soil types (Table 2). Cumulative CO₂ production at the end of incubation (day 90) was also the lowest in the air-dried treatments for all the three soils (Figure 2). However, across the moisture levels, cumulative CO₂ production showed a bell-shaped second-order polynomial response with increased soil moisture was observed for Ultisol and Alfisol soils (Figures 2A,B), and an increasing second-order polynomial response for the Vertisol soil (Figure 2C). The second order polynomial curves were made using the actual soil moisture values, which are continuous variable. For Ultisol, soil respiration varied from 89 μg C g⁻¹ dry soil to 265 μg C g⁻¹ dry soil with the lowest respiration observed in air-dried treatment and the highest at 50% WHC (Figure 2A). Similarly, for Alfisol, the cumulative CO₂ production was the highest in the 100% WHC treatment which was over 200% higher than the CO₂ production in the air-dried treatment (Figure 2B). The cumulative CO₂ production from the 175% WHC treatment also averaged 17% lower than that at 100% WHC. In the Vertisol, the cumulative CO₂ increased with increased moisture (Figure 2C). Multiple linear regression of microbial respiration to soil properties revealed that respiration was significantly dependent on clay content, soil moisture, and SOC content (Table 3).

Microbial respiration data was modeled by a double pool exponential decay model to determine SOC fractions that are actively decomposing (active pool, C1) as well as relatively slowly decomposing (slow pool, C2) at different moisture levels. Regardless of soil texture and soil moisture differences, < 1.5% of the initial SOC respired was partitioned as C₁ (Table 4). For both loamy and clayey soils, C₁ decreased with increasing moisture content. However, for Ultisol, C₁ was similar for all moisture levels except the air-dried treatment. The size of C₂ only varied between 99.2 and 99.7% of initial SOC among all soil types and soil moisture levels, and no significant differences were observed among soil moisture treatments.

Mineralization rate, k₁ of C₁ pool was the highest at 50% WHC (0.23 days⁻¹) for Ultisol, at 100% WHC (0.21 days⁻¹) and at 175% WHC (0.23 days⁻¹) moisture treatments for Alfisol, and at 175% WHC moisture content (0.17 days⁻¹) for Vertisol (Figure 3A). Mineralization rate, k₂, of C₂ pool was one to two order of magnitude lower than k₁ and showed no significant differences across moisture levels in all the three soils (Figure 3B). Comparing the soil types, Ultisol had comparatively higher k₂ followed by Alfisol and Vertisol.

The EOC values were generally higher in the drier treatments compared to the wetter treatments (Figure 4). For Ultisol, EOC was statistically influenced by moisture (p < 0.001), length of incubation (p < 0.0001), and their interaction (p < 0.001) while for Alfisol and Ultisol soils, moisture and length of incubation significantly influenced the EOC and the interaction between moisture and length of incubation was not significant (p > 0.05). The EOC concentration at day 1 for all moisture levels and soil types was significantly higher (60–115 mg kg⁻¹) than that at days 15 (33–103 mg kg⁻¹), 60 (24–94 mg kg⁻¹), and 90 (22–89 mg kg⁻¹). As incubation progressed, for Ultisol and Alfisol soils, we observed no significant differences in EOC content of the driest treatment with time but for other treatments EOC decreased with time. However, for Vertisol, EOC in the driest treatment also decreased with incubation length. More importantly, beginning day 15, the lowest EOC content in each soil type was recorded from the moisture treatment which produced the highest cumulative respiration: at 50, 100, and 175% WHC for Ultisol, Alfisol, and Vertisol soils, respectively, (Figures 2, 4).

There was a consistent increase in total C acquisition enzyme activities (sum of BG, AG, CBH, and XYL) with increase in soil moisture. The highest enzyme activity was observed at the highest soil moisture content for all the three soils at all time points (Figure 5). Also, regardless of soil types, air-dried treatments consistently showed the lowest enzyme activity and there was a decrease in enzyme activity with the length of incubation. For Ultisol, on day 1, we observed a significant increase in enzyme activity with an increase in moisture treatments but on day 90, no statistical difference in enzyme activity was observed in the moist treatments (only air-dried treatment showed lowest activity). The regression analysis showed a significant positive relationship between soil moisture and enzyme activity (Table 5). However, the PERMANOVA analysis revealed that soil moisture treatments did not influence the microbial community composition derived by PLFA in any soil type (p = 0.35 for Ultisol, 0.41 for Alfisol, and 0.12 for Vertisol) (Supplementary Table S2). Our Mantel test results showed that the microbial community composition did not influence the enzyme activity (p > 0.05) (Supplementary Table S1).