Original date

Prior to the autumn maize harvest in 2023, composite topsoil samples (0–20 cm depth) were obtained from each plot by mixing five 5-cm diameter cores collected in a S-shaped pattern, with three replicates per treatment. All samples were immediately taken back to the laboratory on dry ice. Visible plant/animal residues and gravel were carefully removed before each sample was divided into two parts: one for bulk soil analysis, and the other for aggregate fractionation. Aggregates were separated by the Savinov dry-sieving method to minimize disruption of non-water-stable aggregates and retain water-soluble substances . In brief, soil clods were gently broken along natural planes and passed through an 8-mm sieve, then air-dried under shaded. Subsequently, 60 g of air-dried soil was placed on the top sieve of a nested set (2 mm, 0.25 mm, and a collection pan) and shaken for 3 minutes using a mechanical shaker (amplitude: 2 mm, frequency: 50 Hz). Aggregates retained on each sieve and the pan were carefully collected and weighed to obtain large aggregates (LA, > 2 mm), small aggregates (SA, 0.25–2 mm), and micro-aggregates (MA, 10 assessment.Various forms of iron (Fe) and aluminum (Al) oxides were measured to assess their potential mediation of SOC persistence within aggregates. Free Fe and Al oxides (Fed + Ald) were measured using a sodium dithionite, sodium citrate, and sodium bicarbonate extraction method. Amorphous Fe and Al oxides (Feo + Alo) were extracted with acid ammonium oxalate buffer. Complex Fe and Al oxides (Fep + Alp) were extracted using sodium pyrophosphate . Oxide contents were expressed as 0.5Fe + Al.The chemical stability (CS) of soil organic matter was evaluated by measuring aggregate-associated organic compounds via pyrolysis gas chromatography-mass spectrometry (Py-GC/MS). Briefly, 10 mg of aggregate sample was mixed with 1 ml of 25% tetramethylammonium hydroxide in a quartz boat and dried under N2 flow. The dried residue was transferred to a Pyrex tube reactor (EGA/PY-303D) and pyrolyzed at 400°C for 5 min. The pyrolysis products were trapped in chloroform (50 ml, ice bath), concentrated by rotary evaporation, and re-dissolved in 0.2 ml chloroform for analysis on a GCMS-QP2010 Ultra (Shimadzu, Japan).Organic compounds were identified and quantified using the Automated Mass Spectral Deconvolution and Identification System and the NIST mass spectral library . The CS was calculated using the following equation: CS=OCR/OCL, where, OCR is the total concentration of recalcitrant organic compounds (phenols, aromatic hydrocarbons, carboxylic acids, and amides), and OCL is the total concentration of labile organic compounds (alcohols, lipids, ethers, and other aliphatics).Microbial activity in aggregates was assessed by measuring the activities of three extracellular enzymes involved in C cycling: α-glucosidase (AG), β-glucosidase (BG), and β-cellobiohydrolase (CBH). Briefly, 1 g of pre-incubated sample was mixed with 100 ml sodium acetate buffer to prepare a soil suspension. A 96-well microplate was set up with eight replicates of each of the following well types: 200 μl suspension + 50 μl buffer, 50 μl substrate (Table S1) + 200 μl buffer, 50 μl standard + 200 μl suspension, and 50 μl standard + 200 μl buffer. After incubation in the dark at 25°C for 4 h, the reaction was terminated by adding 10 µl of 1.0 mol L–1 NaOH to each well. Fluorescence was measured at 365 nm excitation and 450 nm emission using a microplate reader (Synergy H1M, USA).Bacterial and fungal community composition was assessed by high-throughput sequencing (Illumina MiSeq PE300, USA). The bacterial 16S rRNA gene was amplified using primers 515F (5′-GTGCCAGCMGCCGCGG-3′) and 907R (5′-CCGTCAATTCMTT TRAGTTT-3′), and the fungal ITS region was amplified with primers ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA -3′) and ITS2R (5′-GCTGCGTTCTTCATCGATGC-3′). Microbial diversity indices (e.g., Shannon index) were calculated from OTU tables in Mothur (V1.30.2, https://www.mothur.org/ wiki/Download_mothur). Taxonomic classification of bacteria and fungi at class level was performed using the SILVA (release 138) and UNITE (release 8.0) databases, respectively.