Dynamics of bacterial and archaeal communities during horse bedding and green waste composting

Study site, sampling, and physicochemical analyses

Three windrows (22 m × 5 m × 3 m) containing horse bedding (wood chips and horse manure) and green plant residues of varying maturity were sampled in summer 2018 (average site temperature = 27 °C). Residues in the youngest windrow were between 1 and 6 weeks old (Young), those in the second windrow were 3 months old (Middle), and those in the oldest windrow were 12 months old (Aged). All three windrows had been mixed with a tractor 2-3 times per month since they were put in place. Fresh horse bedding (Litter), the Young, Middle, and Aged piles, and a 24-month-old mature compost pile (Compost) were all sampled for analysis (Fig. 1) (Grenier, 2021). The three windrows were divided lengthwise into four sections and each section was sampled four times, taking two samples from each side of the windrow, while temperature was also measured at all 16 sampling points. Approximately 1 kg of material was collected at a depth of 60 cm at each sampling point, and the 4 samples from the same section were pooled to generate 4 composite samples per windrow (n = 4). As described in Grenier (2021), four composite samples were collected at similar depths in the horse bedding pile and the mature compost pile. Samples were split and fractions were either frozen at −80 °C, refrigerated (4 °C), oven dried at 105 °C for 24 h, or air dried for two weeks. The oven- and air-dried samples were ground to a particle size of <2 mm before being stored in the dark until analysis. Organic matter was determined on the oven-dried samples by determining the weight loss after ignition at 600 °C. The oven-dried samples were used for analysis of the total carbon (C) and nitrogen (N) content by dry combustion at 950 °C using the varioMICROcube analyzer (Elementar, Langenselbold, Germany) and then extracted with water (1:10 (w/v)) for pH determination. Total mineral nitrogen (NH₄⁺ and NO₂⁻-NO₃⁻) was measured on fresh samples extracted with 2.0M KCl (Carter & Gregorich, 2006) using methods for soil samples with the QuikChem^® 8500 Series 2 FIA system (Lachat Instruments, Milwaukee, WI).

Cellulose, hemicellulose and lignin content were determined on air-dried samples with the ANKOM2000 Automated Fiber Analyzer (Ankom Tech., Macedon, NY, USA) according to the manufacturer’s guidelines (http://www.ankom.com/procedures.aspx). Hemicellulose content was estimated as the difference between the neutral-detergent fiber (NDF) and the acid-detergent fiber (ADF), cellulose content was estimated as the difference between the ADF and the acid-detergent lignin (ADL) and lignin content was estimated as the difference between the ADL and the ash content.

DNA extraction, PCR amplification, sequencing and processing

Total genomic DNA was extracted (250 mg of frozen samples) using Qiagen’s DNeasy Power Soil^® Pro kit and its quantity and quality were later determined spectrophotometrically using Thermo Fisher’s NanoDrop 2000c. PCR reactions were performed using the forward primer P609D (5′- GGMTTAGATACCCBDGTA- 3′) and reverse primer P699R (5′- GGGTYKCGCTCGTTR-3′) targeting the V5-V6 region of the 16S ribosomal RNA (rRNA) gene and providing excellent coverage for bacterial and archaeal species (Klindworth et al., 2013). The amplification was performed under the following conditions: initial denaturation 94 °C for 2 min, denaturation 94 °C for 30 s, annealing 58 °C for 30 s, extension 72 °C for 30 s, final extension 72 °C for 7 min, 4 °C hold, over 35 cycles. The resulting amplicons (± 329 bp) were sequenced via Illumina MiSeq 2500 paired end 2 X 250 pb platform at the Genome Quebec Innovation Centre (Montreal, Canada). Reagent controls for quality assurance were below the detection limit.

The ANCHOR pipeline (Gonzalez, Pitre & Brereton, 2019) was used for the processing and annotation of sequence reads (https://github.com/gonzalezem/ANCHOR_v1.0). First, sequences were aligned and dereplicated using Mothur, then Exact Sequence Variants (ESVs) were selected based on a count threshold of 12 across all samples (n = 20). The annotation was performed with strict BLASTn criteria (99% identity and coverage) on 4 sequence repositories: NCBI-curated bacterial and Archaea RefSeq, NCBI nr/nt, SILVA and Ribosomal Database Project (RDP) (NCBI-curated bacterial and Archaea RefSeq is given a priority when at 100% identity and coverage). An Ambiguous hit annotation was retained and reported when multiple, equally good (highest identity/coverage), annotation was found. Amplicons with fewer than 12 counts across all samples were binned to high-count sequences in a second BLASTn, using a threshold of 98% identity/coverage. As databases are subjected to changes and updates, all annotations should be considered as assumptions and interpreted with caution.

Statistical and differential abundance analysis

All statistical analysis for the physicochemical measurements were carried out in GraphPad Prism 8.4.3. The physicochemical data were tested for normal distribution using Shapiro–Wilk test (File S1) and the parameters that failed the test (NO₂⁻-NO₃⁻) underwent a square root transformation. One-way ANOVA followed by a Tukey’s multiple comparisons post hoc test was used to compare physicochemical properties across successive composting phases. A Spearman correlation matrix presenting the interactions between physicochemical parameters (temperature, O.M., pH, total carbon, cellulose, hemicellulose and lignin content, total nitrogen NH₄⁺ and NO₂⁻-NO₃⁻ content) and the 50 ESVs with the highest relative abundance over all five composting phases was produced to demonstrate the relationship between physicochemical parameters and microbial taxa dynamics. Alpha diversity was measured using Shannon index indices within Phyloseq package (McMurdie & Holmes, 2013). Alpha diversity was compared between the different groups of samples using a t-test. The Phyloseq package (McMurdie & Holmes, 2013) was used to perform a Redundancy Analysis (RDA) ordination based on Bray-Curtis ecological distances, while the veganCovEllipse function from Vegan package (Oksanen et al., 2008) in R (R Core Team, 2021) was used to draw the dispersion ellipses. Finally, the adonis function in the Vegan R Package was used to perform a PERMANOVA on the Bray distances matrices to evaluate the significant differences between the sampled compost piles. Differential abundance analysis on ESVs was performed using DESeq2 (Love, Huber & Anders, 2014; Thorsen et al., 2016), which was specifically conceived to offer good performance with uneven library sizes and sparsity (Brereton, Pitre & Gonzalez, 2021; Gonzalez, Pitre & Brereton, 2019; Weiss et al., 2017). A false discovery rate (FDR; Benjamini–Hochberg procedure) < 0.05 was applied (Anders et al., 2013; Love, Huber & Anders, 2014). Raw counts were log transformed across samples (rlog function, R Phyloseq package). Sparsity and low-count cut-offs were applied as ESV counts must be > 2 in 40% of the samples (Dhariwal et al., 2017; Gonzalez, Pitre & Brereton, 2019) while ESV count in a single sample is < 90% of the count in all samples.

The interpretation of the results focused on the identification of bacteria potentially involved in the transformation of carbon and nitrogen from lignocellulose. The association of bacterial species with potential functions is reported in the literature (File S2) but should be considered speculative as functions were not measured directly. To increase the confidence potential functional association, only ESVs identified at the species level, without ambiguous annotation and with an identity score of 100% were screened for roles in cellulose and lignin degradation, methane production and oxidation, ammonification, ammonium and nitrite oxidation and denitrification.

Physiochemical properties of the composting windrow

Physicochemical measurements were performed on the horse litter (Litter), three windrows of 1.5, 3 and 12 months of age (Young, Middle and Aged, respectively), and mature compost (Compost) (Fig. 2, raw data and complete statistical results in File S1). The average temperature of the horse litter pile was 38.2 °C. Temperatures varied significantly (p < 0.05) between each successive compost phase, averaging 67.8 °C for Young, 62.1 °C for Middle, 46.1 °C for Aged and 37.0 °C for mature Compost. Organic matter content was 93.1% in the Litter and progressively reduced to 56.1% in Young, 42.9% in Middle, 31.7% in the Aged and 24.4% in the mature Compost. This decrease was significant between Litter and Young phase as well as between Young phase and Compost. The initial pH in the Litter was at 7.37, which increased significantly from pH 7.33 in the Young phase to pH 7.72 in Middle phase, and significantly increased to pH 8.42 in Aged phase and pH 8.23 in the mature compost.

Total carbon was 43.9% in Litter and decreased significantly to 26.9% in Young. Middle, Aged and Compost were similar, averaging 22.9%, 22.0% and 18.5% of carbon, respectively, with Compost being significantly lower than Young (Fig. 2, File S1). Nitrogen levels were at 1.37% in Litter before decreasing significantly to 1.09% in Young phase and remaining similar for Middle, Aged and Compost at 1.08%, 1.10% and 1.04%, respectively. The C:N ratio reduced significantly from 32.6 in the Litter to 22.7 in Young phase and then remained similar at 21.0 in Middle phase, 19.9 in Aged phase and 17.7 in the mature Compost.

Cellulose content dropped significantly from 32.20% (dry matter) in the Litter to 10.86% in the Young phase and then again to 6.19% in Middle phase, before reaching 5.50% in the Aged phase and 3.86% in the mature Compost (but was not significantly different compared to Middle phase) (Fig. 2). The hemicellulose content decreased significantly from Litter to Young, from 16.33% dry matter to 6.94%, and remained similar at 5.75% for Middle, 5.96% for Aged and 5.79% for the mature Compost. Lignin significantly decreased from 17.31% in Litter to 9.87% in Young phase. The lignin content of 7.79% in Middle, 7.99% in Aged and 6.49% in the mature Compost did not vary significantly from Young phase.

Ammonium (NH₄⁺) content increased significantly from 16.45 mg kg⁻¹ in Litter to 31.07 mg kg⁻¹ in Young phase, and then dropped significantly from Young phase to 3.86 mg kg⁻¹in Middle phase. The NH₄⁺ content was 0.87 mg kg⁻¹ in Aged phase and 0.54 mg kg⁻¹ in Compost, significantly lower than Litter and Young phase. The NO₂⁻-NO₃⁻ content was of 10.11 mg kg⁻¹ in Litter, 13.70 mg kg⁻¹ in Young phase, 36.7 mg kg⁻¹ in Middle phase, 24.85 mg kg⁻¹ in Aged phase and 6.57 mg kg⁻¹ in Compost. The NO₂⁻-NO₃⁻ content increased between Litter and Middle phase and decreased significantly between the Middle phase and Compost phases.

Composting microbial community

A total of 3,133,873 amplicons were aligned with lengths (>0.1% counts) ranging 322-362 nt and 2,612 ESVs (Exact Sequenced Variants) were identified (File S2). Of these ESVs, 517 (19.8%) could be annotated as putative species, 694 (26.6%) could be annotated at the genera level, 486 (18.6%) at higher taxonomic levels and 915 (35.0%) were annotated as unknown (Fig. 3B). ESVs identified as putative species captured 32.3% of raw amplicon counts (Fig. 3C), had an average identity of 99.8%, including 186 ambiguous ESVs (similarity to multiple taxa). ESVs annotated at genus level captured 25.4% of raw counts, ESVs annotated at higher taxonomy level captured 25.3%, and unknown ESVs captured 17.0%. ESVs identified across all phases belonged to 26 different phyla, of which Proteobacteria, Firmicutes, Actinobacteria, Bacteroidetes and Chloroflexi represent 56.0% of the total ESV diversity and shared 73.5% of the total raw counts, while archaea represented 21 ESVs and 1.7% of total raw counts (Fig. 3A, stacked bar graph of relative abundance in File S2). Of the 517 ESVs identified to the species level, Thermobifida_fusca_2, Thermomonospora_chromogena_1 and Thermobifida_bifida_1 are the most abundant and account for 5.6% of the raw counts.

A Spearman correlation matrix plotted the physicochemical parameters measured in each phase against the relative abundance of the 50 most abundant species-identified ESVs (Fig. 4A, File S2). The 50 ESVs, sorted according to the phase in which they are most abundant (Fig. 4B), show a strong pattern suggesting that the abundant ESVs within a phase are largely correlated with the same physicochemical parameters. The highly abundant ESVs in Litter were strongly correlated with the amount of organic matter, including total carbon and nitrogen, plant cell wall components (cellulose, hemicellulose and lignin) and ammonium. ESVs were mostly not correlated with temperature and NO₂⁻-NO₃⁻ content while they showed a strong negative correlation with pH. The correlation patterns were similar for ESVs abundant in the Young and Middle phases, which correspond to the thermophilic phases of the process. The relative abundance of ESVs was mainly correlated with temperature and NH₄⁺ content for ESVs abundant in Young and temperature and NO₂⁻-NO₃⁻ content for those abundant in Middle, while it was generally negatively correlated with OM, total carbon and nitrogen, and plant wall components. The correlation patterns were also similar for abundant ESVs in the Aged and Compost phases with a contrasting profile to the one observed in Litter. The relative abundance of ESVs was very negatively correlated with the amount of OM, plant cell wall components, total carbon and nitrogen, and NH₄⁺ content. The eight relevant ESVs were, nevertheless, strongly correlated with pH, which was significantly higher in the Aged and Compost phases (Fig. 2).

Shannon’s alpha-diversity index was significantly different between successive groups (t-test p < 0.05), except for Young phase vs. Middle phase (Fig. 3D) (Observed, Chao1, se.chao1, Shannon, Simpson, InvSimpson and Fisher indexes can be found in File S1). The lowest alpha diversity index was found in Litter and progressively increased at each phase with mature Compost having the highest index. Redundancy analysis (RDA) indicated that samples separated by time, with the first principal coordinate explaining 46.4% of the variance between the samples (Fig. 3E) and multivariate analysis identified significant variance between each phase (PERMANOVA, p < 0.001). A total of 810 ESVs were identified as significantly differently abundant (DESeq2, padj < 0.05) between the Litter and Young phase, 653 between the Young and Middle phases, 1182 between Middle and Aged phases and 663 between Aged phase and Compost (File S2).

Differential abundance between Litter and Young phases

A total of 64 DA ESVs from Proteobacteria, 31 from Actinobacteria, 17 from Bacteroidetes and 15 from Firmicutes were observed as significantly reduced between Litter and Young phases (Fig. 3F), including 11 from the family Microbacteriaceae, seven from Bacillaceae, seven from Xanthomonadaceae and six from Alcaligenaceae. Reductions in the relative abundance of 64 DA ESVs annotated as putative species by a fold change of 5.8–8.0 were observed and included Sphingobacterium_jejuense_1, Delftia_litopenaei_1, Sandaracinus_amylolyticus_1, Nakamurella_panacisegetis_1 and Sphingobacterium_cibi_1 (File S2). A high number of DA ESVs increased in relative abundance from Litter to Young phases including 319 ESVs from Firmicutes, 133 from Actinobacteria, 119 from Proteobacteria and 27 from Chloroflexi. The largest amount of these significantly increased DA ESVs were from Firmicutes families (e.g., Paenibacillaceae, Bacillaceae and Clostridiaceae), but the highest fold change increases included non-Firmicutes such as Thermomicrobium_carboxidum_1, Rhodothermus_profundi_1, Rhodothermus_marinus_1, Thermogutta_terrifontis_1, Thermus_thermophilus_1, Rhodothermus_marinus_2 and Thermoleophilum_album_1, which ranged between 11.0–13.6 fold higher in relative abundance in Young compared to Litter.

Differential abundance between Young and Middle phases

DA ESVs observed as significantly lower in relative abundance in Middle compared to Young were dominated by 80 ESVs from Firmicutes, 55 from Actinobacteria and 35 from Proteobacteria, which included (e.g., Bacillaceae, Microbacteriaceae, Limnochordaceae, Paenibacillaceae and Thermoactinomycetaceae) (Fig. 3F). Large reductions in relative abundance of 101 DA ESVs annotated as putative species included examples such as Jeotgalicoccus_psychrophilus_1, Corynebacterium_stationis_1, Corynebacterium _guangdongense_1, Corynebacterium_casei_1, Streptococcus_equinus_1, Kurthia _massiliensis_1, Weissella_confusa_1 which had reduced relative abundance by a fold change of 7.2−8.4 (File S2). The most highly represented phyla in DA ESVs which significantly increased in relative abundance from Young to Middle phases included 66 ESVs from Proteobacteria, 37 from Firmicutes, 31 from Actinobacteria, 18 from Chloroflexi and 11 from Planctomycetes. The DA ESVs significantly increasing in the Middle phase included examples such as Caenibacillus_caldisaponilyticus_1 and Methylocaldum_szegediense_1 as well as the Verrucomicrobia Limisphaera_ngatamarikiensis_1, which ranged from a fold change increase of 3.8–4.1. A total of six DA ESVs identified as putative archaea were also significantly increased in relative abundance from Young to Middle phase, including Methanothrix_thermoacetophila_1 and Methanoculleus_thermophilus_1 by a fold change of 5.7 and 1.8, respectively.

Differential abundance between Middle and Aged phases

Changes between the Middle and Aged phases included substantial reductions of 151 DA ESVs from the phyla Firmicutes, (Paenibacillaceae and Bacillaceae), as well as 46 from Actinobacteria and 25 from Proteobacteria (Fig. 3F). From the 116 DA ESVs annotated as putative species, examples of substantial shifts included Limisphaera_ngatamarikiensis_1, Rhodothermus_profundi_1, Cohnella_laeviribosi_1, Thermomicrobium_carboxidum_1, Thermogutta_terrifontis_1 and Aneurinibacillus_thermoaerophilus_1, which reduced by 4.9−7.3- fold change (File S2). DA ESVs significantly increased in relative abundance from Middle to Aged phase included 213 ESVs from Proteobacteria, 68 from Actinobacteria, 31 from Chloroflexi, 25 from Acidobacteria, 18 from Planctomycetes, 15 from Bacteroidetes and 15 from Gemmatimonadetes. From the 49 DA EVSs annotated as putative species which significantly increased, large changes included two archaea, Nitrosotenuis_cloacae_1 and Methanosaeta_concilii_1, as well as Methylocaldum_marinum_1, Ignavibacterium_album_1, Syntrophobacter_sulfatireducens_1, Sulfurivermis_fontis_1 and Denitratisoma_oestradiolicum_1, which increase by 5.6–8.8-fold change.

Differential abundance between Aged phase and mature Compost

DA ESVs which significantly reduced in relative abundance between Aged and mature Compost phases were dominated by 65 ESVs from Proteobacteria, 24 from Firmicutes, 22 from Actinobacteria and 17 from Chloroflexi, and were very widely distributed over 100 families, with the most frequent including eight ESVs from Bacillaceae, six from Caldilineaceae, five from Nitrosomonadaceae and five from Paenibacillaceae (Fig. 3F). From the 40 DA ESVs annotated as putative species, examples of substantial changes included Ignavibacterium_album_1, Methanothrix_thermoacetophila_1 (Archaea), Thermogutta_terrifontis_1, Leucobacter_chromiireducens_1, Rhodothermus_profundi_1 and Rhodothermus_marinus_2, which ranged from a reduction of between 4.4–6.5-fold change (File S2). DA ESVs which significantly increased in relative abundance between Aged and mature Compost phases were dominated by 147 ESVs from Proteobacteria, 45 from Actinobacteria, 19 from Bacteroidetes and 11 from Chloroflexi, including 51 ESVs from within the order Rhizobiales, such as Methylocystis_rosea_1, Mesorhizobium_tamadayense_1 and Rhizobium_helanshanense_1 which increased in relative abundance by between 3.7–4.4-fold change. From the 55 DA ESVs annotated as putative species, the largest changes included Methylosarcina_lacus_1, Flavobacterium _degerlachei_1, Lysobacter_yangpyeongensis_2 and Nitrospira_japonica_1, which ranged from a reduction of between 6.9–7.4-fold change.

Microbes associated with carbon dynamics

Lignocellulose degradation

Seventy-two differently abundant ESVs were annotated as species associated with putative cellulose degradation potential (cellulase or β-glucanase activity; Fig. 5, File S2). These included species within Firmicutes (34), Actinobacteria (24), Proteobacteria (6), Bacteroidetes (5), Chloroflexi (2) and Deinoccocus Thermus (1). In Young phase, 48 ESVs annotated as species associated with potential cellulose degradation significantly increased in relative abundance compared to Litter; these were dominated by ESVs from the order Bacillales (22), and included Ammoniphilus_resinae_2, Bacillus_borbori_1, Bacillus_coagulans_1, Brevibacillus_borstelensis_1, Brevibacillus_thermoruber_1 and _2, Cohnella_panacarvi_1, Geobacillus_stearothermophilus_2, Geobacillus_thermocatenulatus_1, Geobacillus_thermodenitrificans_1, _2, _3 and _4, Paenibacillus_barengoltzii_3, Paenibacillus_ginsengihumi_1, Paenibacillus_ihumii_1 and _2, Paenibacillus_lactis_1 and _2, Thermobacillus_composti_1 and _2 and Ureibacillus_terrenus_1 (File S2). In Middle phase, 22 ESVs annotated as species associated with potential cellulose degradation were significantly reduced in relative abundance and only 5 were significantly increased in relative abundance, the Clostridiales, Clostridium_colicanis_1, Gracilibacter_thermotolerans_1 and Ruminiclostridium_thermocellum_1, as well as the Streptosporangiales, Thermomonospora_chromogena_1 and Thermomonospora_curvata_1. Similarly, a further substantial reduction of 33 ESVs significantly reduced from Middle to Aged phase, while only three increased. Aged phase and mature Compost were relatively similar with only seven ESVs annotated as species associated with putative cellulose degradation significantly increasing: Actinotalea_ferrariae_1, Devosia_honganensis_1, Flavobacterium_degerlachei_1, Hyphomicrobium_zavarzinii_1, Lapillicoccus_jejuensis_1, Nocardioides_aestuarii_1 and Sorangium_cellulosum_2.

Thirteen differentially abundant ESVs were annotated as species associated with lignin degradation potential, belonging to the phyla Actinobacteria (5), Proteobacteria (2) and Firmicutes (6) (Fig. 5). Of these, four ESVs were significantly reduced in relative abundance from Litter to Young phase (Comamonas_testosteroni_1, Gordonia_paraffinivorans_1, Mycobacterium _thermoresistible_1 and Rhodococcus_zopfii_1) while seven ESVs significantly increased in relative abundance (Bacillus_benzoevorans_1, Brevibacillus_borstelensis_1, Brevibacillus _thermoruber_1, Brevibacillus_thermoruber_2, Ochrobactrum_intermedium_1, Thermomonospora_curvata_1 and Ureibacillus_terrenus_1). Five ESVs were significantly reduced from Young to Middle phase, while only one ESV was significantly increased, (Thermomonospora_curvata_1). Six ESVs annotated as species associated with lignin degradation potential were differentially abundant between Middle and Aged phases, all of which were significantly reduced in relative abundance in Aged phases, Brevibacillus_borstelensis_1, Gordonia_paraffinivorans_1, Thermobifida_cellulosilytica_1, Thermomonospora_curvata_1, Ureibacillus_terrenus_1 and Ureibacillus_thermosphaericus_1, while no significant changes were identified between Aged phase and mature Compost.

Microbes associated with nitrogen cycling

Physiochemical changes throughout composting

The high recorded temperature during Young (67.8 °C) and Middle (62.1 °C) phases were accompanied by a significant decrease in 54% of the amount of organic matter at the beginning of the composting process (Fig. 2). The substantial changes from Litter to Middle phase suggest that most of the organic matter decomposition occurred within the first three months, under thermophilic conditions. The thermophilic phase is often considered as the most microbially active as the high temperatures increase the reaction rates and the efficiency of thermostable enzymes to reach an optimal level of lignocellulose degradation (Ryckeboer et al., 2003).

The three main plant structural components; cellulose, hemicelluloses and lignin were quantified at each phase to monitor their degradation (Fig. 2). For all three, the sharpest decrease occurred between Litter and Young, with a 57.5% decrease in hemicellulose, 66.3% decrease in cellulose and a 43.0% decrease in lignin. A significant decrease between Young and Middle was observed for cellulose, but not for hemicelluloses and lignin. These results suggest that the majority of lignocellulose degradation occurred during the thermophilic stages, which is consistent with expected temperatures of 55−60 °C for optimal lignocellulolytic activity in composting (Tuomela et al., 2000).

Ammonium increased significantly in Young before a significant decrease was observed in Middle, while NO₂⁻-NO₃⁻ content increased between Litter and Middle phase before decreasing significantly between the Middle phase and Compost (Fig. 2). The increase in NH₄⁺ content corresponds to high reductions in organic matter, here being plant material and horse feces, and is expected during the thermophilic phase (Bernal, Alburquerque & Moral, 2009). The following decreases in NH₄⁺ could be due to uptake by microorganisms, direct oxidation by ammonia oxidizing microorganisms (AOB/AOA), or through loss by volatilization of NH₃. The accumulation of NO₂⁻-NO₃⁻ in the following phases and high reduction in total nitrogen suggests ammonia oxidation as well as volatilization may be occurring. Nitrate content is a major criterion of compost quality (Bernal et al., 1998), so the drop in NO₂⁻-NO₃⁻ concentrations during the later Aged phase and mature Compost could be considered as deleterious to high quality compost production.

Microbial community structure and composition through composting

The most diverse and abundant bacterial phyla; Proteobacteria, Firmicutes, Actinobacteria, Bacteroidetes and Chloroflexi (Fig. 3A), are commonly reported as dominant in compost, ranging from 72% to 92% of microbial diversity (Antunes et al., 2016; Partanen et al., 2010; Wei et al., 2018; Zhou et al., 2018). A total of 299 distinct putative species of bacteria and archaea were identified across the different composting phases. Of these, 50 are shown in Fig. 4. These 50 ESVs are found to have the highest relative abundance across all composting phases and belong to 35 different genera, the majority of which are in the phyla Firmicutes (13), Actinobacteria (eight) and Proteobacteria (eight). Moreover, two of the three species identified here as having highest relative abundance across all samples, Thermobifida fusca and Thermobifida bifida, agree with the dominant genera, Thermobifida and Thermopolyspora, found by Zhang et al. (2015a) and Zhang et al. (2015b) within maize straw compost. This high diversity illustrates the complexity of compost systems, although the substantial amount of remaining unknown or poorly characterized ESVs (Fig. 3B) indicates the extensive amount of natural diversity left to be explored in everyday biological systems, such as compost.

Ordination of samples suggested substantial differences in the microbial community between each composting phase (Fig. 3E). Bacterial diversity increased as composting phases progressed, with few species with high relative abundance in early phases and many species present at a lower abundance in the mature compost (Fig. 3D). Previous research has suggested that limited resources within mature compost can create strong competition and limit diversity of bacteria (Antunes et al., 2016; De Gannes, Eudoxie & Hickey, 2013). However, diversity indices reported here suggest that high temperatures are likely to have created stronger selection and limited diversity during thermophilic phases, as observed elsewhere (Ryckeboer et al., 2003; Zhou et al., 2018). This was further supported during differential abundance (DA) analysis, which revealed that the largest numbers of species changes occurred between Litter and Young, and Middle and Aged phases (Fig. 3F, File S2), corresponding to the largest shifts in temperature. These general changes can be attributed to several factors, but temperature, pH, and OM content, including total carbon and nitrogen, are probably the major drivers of composting microbial community changes, as presented in Fig. 4.

Species reduced from Litter to Young were consistent with the rapid increase in temperatures from 38.2 °C to 67.8 °C, with largest reductions in previously reported mesophilic species such as Sphingobacterium jejuense, Sandaracinus amylolyticus, Nakamurella panacisegetis and Sphingobacterium cibi (Kim, Lee & Lee, 2012; Lai et al., 2016; Mohr et al., 2012; Siddiqi et al., 2016) (File S2). The correlation analysis (Fig. 4) indicates a strong effect of temperature on the relative abundance of the main ESVs during the Young phase, suggesting a selection effect of temperature on the species composition. Reductions in species such as Delftia litopenaei also suggests an analogous dynamic in organisms not captured by the 16S rRNA gene amplification, such as insects, as Delftia are dominant gut symbionts in arthropods common to composts (Morales & Wolff, 2010; Wang et al., 2014; Xie et al., 2012). Expected major increases in extremely thermophilic bacteria also occurred, including Thermomicrobium carboxidum, Rhodothermus profundi, Rhodothermus marinus, Thermogutta terrifontis, Thermus thermophilus and Thermoleophilum album, which were all originally isolated from geothermally heated biofilm, hydrothermal vents or hot springs (Alfredson et al., 1988; King & King, 2014; Marteinsson et al., 2010; Oshima & Imahori, 1974; Slobodkina et al., 2015; Zarflla & Perry, 1984) and some of which have previously been reported in compost systems at the specie or genus level (Antunes et al., 2016; Gladden et al., 2011; Varma, Dhamodharan & Kalamdhad, 2018).

Subsequent reductions of species from Young to Middle phases included further loss of thermosensitive species such as Jeotgalicoccus psychrophilus, as well as animal-associated Corynebacterium sp., Kurthia massiliensis and Streptococcus equinus (Bernard et al., 2010; Roux et al., 2012; Schlegel et al., 2003; Yoon et al., 2003), which likely represents reductions in species associated with horse feces present in Litter. The major species increasing in Middle phase included thermophiles common to compost, such as Caenibacillus caldisaponilyticus (Tsujimoto et al., 2016) and unexpected species such as the thermophilic and alkaliphilic Verrucomicrobia Limisphaera ngatamarikiensis (Carere et al., 2020), but was characterized by significant increases in thermophilic methanogenic archaea, the largest change being observed in Methanothrix thermoacetophila (Kamagata et al., 1992). M. thermoacetophila produces methane under lower oxygen conditions such as those of progressing composting of Middle phase and which likely led to the subsequent increases in moderately thermophilic methanotrophs such Methylocaldum szegediense (Medvedkova et al., 2009).

The transition from Middle to Aged phases saw the greatest shift in differentially abundant taxa as well as an increase in pH and substantial decrease in temperature, with a drop from 62.1 °C to 46.1 °C. These physicochemical variations most likely played a role in microbial proliferation, whereas the abundant ESVs in Aged were strongly correlated with pH and had a weak negative correlation with temperature (Fig. 4). This led to subsequent large reductions in the thermophilic species which had increased during Young and Middle phases, including T. carboxidum, R. profundi, T. terrifontis and L. ngatamarikiensis, as well as reductions in thermophilic archaea such as Methanoculleus thermophilus (File S2 and Supplementary DA figures). The significant reduction in L. ngatamarikiensis suggests that the mesophilic Aged phase is likely highly competitive despite reaching an optimal alkaline condition for L. ngatamarikiensis of pH 8.42 (Carere et al., 2020), and illustrates how specific species can be highly transient throughout each composting phase as the ESV Limisphaera_ngatamarikiensis_1 was absent (below detection) in Litter and mature Compost samples. Species which substantially increased in the Aged phase included mesophilic archaea, potentially replacing lost thermophilic archaea in similar niches, such as methanogen Methanosaeta concilii and the ammonia oxidizing Nitrosotenuis cloacae (Li et al., 2016; Patel & Sprott, 1990), which corresponds to the reduced ammonia and increased nitrites/nitrates in middle and Aged phases. Species substantially increasing also included the metabolically flexible Ignavibacterium album (Iino et al., 2010) which, as a generalist, could be benefiting from the extreme disturbance associated with the transition from a thermophilic to mesophilic habitat (Chen et al., 2021). The cross-feeding or syntrophic archaea Methylocaldum marinum and bacteria Syntrophobacter sulfatireducens and Sulfurivermis fontis (Chen, Liu & Dong, 2005; Plugge et al., 2011; Takeuchi et al., 2019) also substantially increased, illustrating the potential advantage provided to cooperative strategies in highly diverse and competitive environments such as Aged phase (Hibbing et al., 2009).

Further reductions in thermophilic archaea and bacteria occurred between Aged phase and mature Compost, including M. thermoacetophila, T. terrifontis, R. profundi and R. marinus as well as the transient Ignavibacterium album and the extremophile and nematode pathogen Leucobacter chromiireducens (Muir & Tan, 2008). Within highly diverse mature compost, large increases were observed in specialized species such as the obligate methanotroph Methylosarcina lacus and the nitrite-oxidizing bacteria Nitrospira japonica (Fujitani et al., 2020; Kalyuzhnaya et al., 2005), but the community was best characterized by increases in 51 ESVs annotated as taxa within the order Rhizobiales which increased when compared to Aged phase, such as the methanotroph Methylocystis rosea, the nodule associated Mesorhizobium tamadayense and Rhizobium helanshanense, suggestive of a compost community which could benefit soil health and plant rhizosphere associations (Qin et al., 2012; Rahalkar et al., 2018; Ramírez-Bahena et al., 2012).

Microbes associated with carbon dynamics

Methanogens and Methylotrophs community

Differentially abundant species identified as methanogens were present throughout the composting phases and belonged to the archaeal families Methanosarcinaceae, Methanomicrobiaceae, Methanocellaceae, Methanobacteriaceae and Methanosaetaceae. The Young and Middle phases hosted both prevalent thermophilic aceticlastic methanogens such as M. thermophilus and M. thermoacetophila (Kamagata et al., 1992; Tian, Wang & Dong, 2010) and thermophilic hydrogenotrophic methanogens, such as Methanocella arvoryzae (Sakai et al., 2010) as well as the hydrogenotrophic methanogen Methanoculleus hydrogenitrophicus, where previously reported isolates were considered as mesophilic (Tian, Wang & Dong, 2010) (Fig. 5, File S2). As temperatures dropped in Aged phase, mesophilic methanogens with characterized strains having optimal growth temperatures of 37−45 °C, such as Methanobacterium formicicum and M. concilii increased in relative abundance (Bryant & Boone, 1987; Patel & Sprott, 1990). This shift from thermophilic to mesophilic methanogens suggests replacement of species within the methanogen functional niche linked to temperature, but could also be driven by lignocellulose substrate availability, due to characterized syntrophic interactions between methanogen community members, particularly within Methanosarcina, and cellulose-degrading bacteria (Conrad, 2020; Lu et al., 2017), concentrated within the Young phase. This hypothesis is supported by a strong negative correlation between OM, total carbon and nitrogen and plant tissue constituents with ESV abundance in the Aged and Compost phases (Fig. 4). The bacteria are thus likely to be more dependent on the compounds resulting from lignocellulose decomposition than on the lignocellulose itself.

ESVs annotated as methylotroph species from Methylococcaceae, Methylocystaceae and Rhodobacteraceae (all within proteobacteria) were differentially abundant through the composting phases. M. szegediense and Methylocaldum tepidum significantly increased in Young phase, both of which are considered thermophilic (Cvejic et al., 2000) and coinciding with increases in methanogens providing substrate (Fig. 5). However, the largest increases in methylotroph species were observed as temperature decreased from Middle to Aged phase, including increases of Methylocaldum marinum and Methylosinus trichosporium, and further increases in M. tepidum. As compost cooled and matured, Methylobacter marinus and M. lacus increased in relative abundance as well as the Rhizobiales species M. rosea and Methylocystis echinoides. These methanotrophs are thought to be environmentally sensitive, with species from Methylocaldum being replaced by species from Methylosinus, Methylobacter, and Methylocystis, as conditions move from thermophilic to mesophilic temperatures (Halet, Boon & Verstraete, 2006). Although the dynamics of methanogens (Thummes, Kämpfer & Jäckel, 2007) and methanotrophs (Chen et al., 2014) have been studied in compost environments, reports of the co-occurrence of methanogenic and methanotrophic species throughout different composting phases is not common. Syntrophic interactions have been characterized as involving consortia of methanogens, such as M. formicicum, with sulphur-reducing bacteria, such as Desulfotomaculum peckii (Song et al., 2019), which significantly increased in the Young phase, and S. sulfaliredcuens (Ahlert et al., 2016; Knittel & Boetius, 2009), which significantly increased in later composting phases. Understanding these dynamics is important for predicting the production and oxidation of methane in compost, which can substantially influence greenhouse gas emissions from composting.

Microbes associated with nitrogen dynamics

Mature compost as a microbial reservoir

Although the microbial community was very dynamic between the early composting phases, only a small number of differentially abundant ESVs were observed between Aged and Compost (Figs. 3F and 5). Thus, there was relative stability in this very diverse community that was also reflected in their relative closeness in ordination (Fig. 3E), regardless of the greater temporal gap between these phases (12 months). The mature compost community included species which could provide beneficial functions to agriculture, such as nitrogen fixation and lignocellulose degradation (Fig. 6). Agriculture may account for 25% of global methane emissions, which were about 145 Mt CH₄ y⁻¹ per year in 2017 (Smith, Reay & Smith, 2021). Agricultural waste composting has been extensively explored for mitigation of methane emissions associated with organic decomposition (Lou & Nair, 2009) due to putative methanotrophic microbes present within composting processes. The use of mature compost rich in methanotrophic bacteria, such as M. lacus, M. marinus, M. echinoides and M. rosea, which substantially increased within mature compost here, could provide a benefit by extending this potential community function to increase the soil methane sink or reduce soil methane emissions after the application in agriculture (Singh et al., 2010). Any such methane reductions associated with agriculture could help to meet the ambitious target set out in the COP26 Global Methane Pledge of a 30% reduction in global methane emissions (compared to 2020 levels). Similarly, in addition to the more general evidence that compost application can help to suppress crop pathogens (Bonanomi et al., 2007; Bonilla et al., 2012; Hoitink & Fahy, 1986; Termorshuizen et al., 2006), some species detected in the mature Compost are considered to be plant growth-promoting bacteria. These include A. ferrariae, Geobacter thiogenes, Geobacillus thermodenitrificans, Geobacillus stearothermophilus, L. chromiireducens, M. tamadayenses, P. aeruginosa, Paenibacillus yonginensis and R. helanshanense, which have been shown to bestow improved crop nutrient acquisition and/or resistance to different abiotic stresses such as drought, salinity, hydrocarbons, heavy metals and herbicides (Aguiar et al., 2020; Marchant & Banat, 2010; Morais et al., 2004; Nevin et al., 2007; Pieterse et al., 2014; Qin et al., 2012; Rahalkar et al., 2018; Ramírez-Bahena et al., 2012; Sukweenadhi et al., 2014). Although species-level resolved profiling of complex microbial communities is challenging, identification and tracking of these potentially beneficial species to crops could inform our understanding of how compost could improve agricultural soils or the environmental impact of agriculture, in addition to the nutrient and soil stability properties traditionally associated with compost application.