In the GET system, biogas production and methane concentration increased markedly immediately after the 2nd addition of rice straw,as our previous research (Chen et al., 2020). To resolve the microbial and metabolic processes underlying this characteristic response, soil samples were collected at 3–5-day intervals. This sampling strategy was designed to capture short-term fluctuations in volatile fatty acids (VFAs) and corresponding shifts in microbial community structure during the early stages of substrate turnover (Figure 1).

The next-generation sequencing (NGS) based on DNA extracted from soil samples was used to observe the microbial community structure involved in rice straw decomposition and methane fermentation in GET systems, which provides a comprehensive overview of taxonomic structure across sampling points. However, because DNA-based NGS profiles include both active and inactive populations, additional approaches were required to evaluate metabolically active microorganisms and to validate temporal changes suggested by high-throughput sequencing.

Therefore, denaturing gradient gel electrophoresis (DGGE) was employed as a complementary method to visualize dominant and dynamically changing microbial populations across sampling points. DGGE enables direct comparison of banding patterns over time and serves as an effective tool for confirming major shifts in community structure, particularly for targeted microbial groups, prior to quantitative analysis.

To further assess microbial activity, total RNA was extracted from the same soil samples and reverse-transcribed into cDNA. Real-time quantitative PCR (qPCR) using specific primers was then performed to quantify temporal changes in RNA-derived gene copy numbers of selected taxa and functional groups. RNA-based qPCR allowed transcriptionally active populations to be distinguished from dormant or residual DNA, enabling interpretation of gene copy number dynamics as indicators of microbial activity rather than mere presence.

By integrating VFA profiles with NGS-based community structure, DGGE-based pattern validation, and RNA-based qPCR activity data, this study established a multi-layered analytical framework to evaluate the reason why the efficiency of the GET system differed substantially following the 2nd rice straw addition (Supplementary Table S2).

The construction method of GET system was mentioned in our previous paper (Chen et al., 2020). In briefly (Figure 1), cut the sun-dried rice straw into approximately 15 cm lengths and mixed it with the paddy soil on the paddy field, then constructed a gas sampling bed (4.5 m in length, 0.8 m in width, and 0.2 m in height) consisting of a gas production bed (L: 3.0 m × W: 0.8 m × H: 0.2 m) and a soil sampling bed (L: 1.0 m × W: 0.8 m × H: 0.2 m) using rotary tiller. The pH and oxidation-reduction potential (Eh) meters (Orion 3-Star plus, Thermo Fisher Scientific, Waltham, MA, United States) and the soil pore water collector, Mizutor (DIK-8392, Daiki Rika Kogyo, Saitama, Japan) were installed at the same height as the ground in the center of the gas sampling beds. The gas and soil sampling beds (i.e., fermentation beds) were covered with an impermeable rubber sheet (10 m in length × 2.0 m in width × 1 mm in thickness) (Taiyo Kogyo, Osaka, Japan), and sealed the edges with soil. After connecting the gas collection apparatus through the interface on the top of the rubber sheet, the experimental area was flooded to make anaerobic fermentation beds and to seal the gas sampling beds using water pressure. The experiments were performed with three beds.

Three additions of rice straw at intervals of 2 months were performed as follows: the 1st addition was 14 kg/m², the 2nd addition was 14 kg/m², and the third addition was 7 kg/m² (Figure 1). Samples for analysis of biogas volume, methane concentration in biogas, and concentration of VFAs were taken at intervals of 3–5 days after rice straw addition. Biogas, soil, and VFAs samples were collected from each of three fermentation beds operated in parallel.

Soil samples from F_1 to F_10 (Figure 1) were collected based on the dynamics of biogas volume, and concentrations of VFAs and methane.

Soil pore water samples (2 mL) were collected from the gas sampling beds using a soil pore water collector. To purify the water samples, an equal volume of chloroform was added to the water samples, followed by vortex mixing (20 s × 3 with 5 s intervals) and centrifugation (3,000 rpm, 15 min, 25 °C). The aqueous supernatant was recovered and filtered through a 0.2 μm PTFE membrane filter (DISMIC-13CP, Advantec Toyo Kaisha, Tokyo, Japan).

Volatile fatty acids were quantified using a high-performance liquid chromatography (HPLC) organic acid analysis system equipped with an electrical conductivity detector (Prominence, Shimadzu, Kyoto, Japan). Separation was performed on two serially connected Shim-pack SCR-102H columns (300 mm × 8.0 mm i.d.) with a Shim-pack SCR-102H(G) guard column (50 mm × 7.8 mm i.d.) maintained at 40 °C. The injection volume was 50 μL and the autosampler was set to 4 °C. The mobile phase consisted of (A) 5 mM sodium p-toluenesulfonate and (B) 5 mM sodium p-toluenesulfonate containing 20 mM Bis-Tris and 100 μM EDTA, delivered at 0.8 mL/min for both pumps (A and B). The total run (analysis) time was 55 min.

Quantification was performed by the external standard method using mixed organic acid standards (pyruvic acid, glyoxylic acid, lactic acid, acetic acid, propionic acid, butyric acid, isovaleric acid, succinic acid, and formic acid), and concentrations were reported as mg/L. Detailed instrument settings are provided in Supplementary Table S1.

DNA and RNA were co-extracted from soil using the FastDNA™ SPIN Kit for Soil (MP Biomedicals; catalog no. 6560200) following the manufacturer’s protocol, with minor modifications as described by Tournier et al. (2015). Briefly, 250 mg of soil (instead of 500 mg recommended by the kit) was placed in a Lysing Matrix E tube and pre-frozen at −80°C overnight. Cells were lysed by bead beating (two cycles, 40 s each at 6.0, with cooling on ice between cycles), and the lysate was clarified by centrifugation. DNA was purified from the Binding Matrix fraction according to the kit instructions and eluted in 150 μL of DES. The corresponding supernatant was used for RNA purification following the kit-based procedure (including alcohol precipitation and RNase-free handling), and RNA was finally resuspended in RNase-free water.

The cDNA was synthesized using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, United States) according to the manufacturer’s protocol.

PCR amplification for DGGE targeting Clostridium was performed in 50 μL reactions using KOD One^® PCR Master Mix (TOYOBO, Osaka, Japan) with primers Chis150f (5′-AAAGGRAGATTAATACCGCATAA-3′) carrying a 5′ GC-clamp and ClostIr (5′-TTCTTCCTAATCTCTACGCA-3′). Thermal cycling conditions were: 98 °C for 5 min; 35 cycles of 98 °C for 5 s, 62 °C for 5 s, and 72 °C for 10 s; followed by 72 °C for 3 min. Amplicon size and specificity were confirmed by electrophoresis of 5 μL PCR product on 2% (w/v) agarose gels.

Denaturing gradient gel electrophoresis (DGGE) was carried out using a DCode Universal Mutation Detection System (Bio-Rad Laboratories, Hercules, CA, USA). PCR products (50 ng per lane) were separated on 10% (w/v) polyacrylamide gels containing a 10–60% denaturing gradient for Clostridium. Then, 100% denaturant was defined as 7 M urea and 40% (v/v) formamide. Gradients were generated using a Model 475 Gradient Delivery System (Bio-Rad). Electrophoresis was performed in 1 × TAE buffer at 58 °C with a pre-run at 160 V for 10 min, followed by electrophoresis at 160 V for 12 h for Clostridium DGGE profiles.

After electrophoresis, gels were stained with SYBR Gold (Thermo Fisher Scientific) for 30 min with gentle agitation and visualized under UV illumination. Dominant bands of interest were excised, and DNA was eluted overnight in 50 μL sterile water. The recovered DNA was re-amplified in 20 μL reactions using KOD One^® PCR Master Mix with primers 150f (5′-AAAGGRAGATTAATACCGCATAA-3′) and ClostIr (5′-TTCTTCCTAATCTCTACGCA-3′) under the following cycling conditions: 98 °C for 5 min; 30 cycles of 98 °C for 5 s, 61.5 °C for 5 s, and 72 °C for 10 s; followed by 72 °C for 3 min.

PCR products were purified using the QIAquick PCR Purification Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions. Purified amplicons were outsourced for sequencing. Purified samples were sent to the analysis company for sequencing. The determined nucleotide sequence was searched for homology by NCBI BLAST.

Extracted genomic DNA was used for Illumina MiSeq library preparation and paired-end sequencing. The V3–V4 region of the 16S rRNA gene was amplified using primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) to characterize microbial community structure and diversity in the GET system.

Raw reads were quality-filtered using Trimmomatic v0.38 with a sliding-window approach; bases were trimmed when the mean Phred score within the window fell below Q20, and reads shorter than 50 bp after trimming were discarded. Adapter and primer sequences were removed using Cutadapt v1.16.

Denoising and amplicon sequence variant (ASV) inference were performed using DADA2 implemented in QIIME2 (QIIME2 v2020.8.0; DADA2 v2020.8.0). Quality filtering was conducted with maxN = 0 and expected error thresholds of maxEE = 2 (forward reads) and maxEE = 3 (reverse reads). Error models were learned from the data using the learnErrors function, followed by dereplication and ASV inference. Paired-end reads were merged with a minimum overlap of 12 bp, requiring identical sequences across the overlap region, and chimeric sequences were removed using the consensus method.

Taxonomic classification of representative ASV sequences was performed using mothur (v1.39.5; classify.seqs) against the Greengenes and RDP reference databases with a confidence threshold of 0.6. ASVs without taxonomic assignments were removed. In addition, ASVs assigned to non-target groups (e.g., archaeal assignments in bacterial 16S datasets) were excluded prior to downstream analyses.

Purified samples were sent to the analysis company for sequencing. The determined nucleotide sequence was searched for homology by NCBI BLAST.

Real-time PCR was performed by Step One Plus™ (Applied Biosystem, Thermo Fisher Scientific) to quantify gene transcription. The sequences of the target genes and the primers for amplifying them were shown in Supplementary Table S2. The reaction composition and reaction conditions of real-time PCR were shown in Supplementary Table S3. For the preparation of the calibration curve, the plasmid DNA prepared in the preparation of the standard fragment and the genomic DNA of the C. butyricum Prazmowski 1880 strain and the B. cereus strain CCM2010 were used. All experiments were performed in triplicates, and the results were expressed as mean ± standard deviation.

Volatile fatty acid (VFA) concentrations are widely recognized as a key parameter for evaluating anaerobic digestion processes (Magdalena et al., 2019). Acetate, in particular, is a crucial substrate for methane production, influencing the proliferation and metabolic function of methanogenic microbial communities. Therefore, this study first performed quantitative analysis of VFAs and examined the overall structural evolution of the microbial community based on sequence data generated through the aforementioned experimental protocol.

In the GET system, three VFAs were mainly detected (Figure 2), i.e., acetic acid, propionic acid, and butyric acid. Within 5 days of the start of the experiment concentrations of acetic acid and butyric acid increased sharply and reached the maximum of 75 and 15 mM, respectively, on day 9 (F_2). Their concentrations then rapidly decreased, reaching nearly zero by day 22 and becoming undetectable on day 40 (F_5). Conversely, propionic acid displayed a more gradual increase, reaching its highest concentration (20 mM) on day 27 (F_4). Although propionic acid persisted for the longest period among the four acids detected, propionic acid also mirrored the decline from day 30, plummeting to zero on day 40 (F_5).

Following the additional batch of rice straw, no VFAs were detected although the biogas production rate was significantly increased, and the concentration of methane reached 60% on day 19 (F_3) and remained constant thereafter despite no detectable VFAs (Figure 2). This indicated that the rice straw was quickly hydrolyzed to monosaccharides such as glucose, and the resultant monosaccharides were converted to VFAs, which were then rapidly utilized as substrates for methane production without their accumulation.

Since the VFAs are essential intermediates for methanogenesis in the anaerobic fermentation process, therefore, our results encourage elucidation of the microbial community structure in the GET system.