How diverse and expressed are biosynthetic gene clusters in a peatland microbiome?

Potential DNA versus RNA actually transcribed

The question is not how many BGCs exist but how many are used. A BGC is a physically clustered set of genes that together build one secondary metabolite; finding one in DNA shows capacity, not activity. To separate the two, the authors recovered BGCs from metagenomes (community DNA) and measured BGC transcription from metatranscriptomes (community RNA) taken from the same peat layers.

The site was the Schlöppnerbrunnen fen in northern Bavaria, a slightly acidic peatland (pH around 4.7 in the top 30 cm) fed by anoxic, iron-rich groundwater. Two parallel cores collected in July 2020 were sectioned at 5 cm intervals. The analysed interval ran from 5 to 30 cm, spanning the shift from the oxic surface layer to anoxic, water-saturated peat. The top 0–5 cm was excluded because field conditions were very dry, and RNA could not be recovered from the 20–30 cm layers despite repeated attempts, so the deepest interval contributes DNA-based but not expression data.

Nearly 10,000 clusters, most without close relatives

BGCs were called with antiSMASH on assembled contigs longer than 5 kb, yielding 9,694 clusters, of which 1,594 (16.4%) were complete. The average cluster length was 31,073 bp. Because short-read assembly can split large modular clusters and inflate counts, the authors flagged incomplete NRPS, PKS and NRPS-PKS hybrid BGCs whose biosynthetic core genes all sat within 5 kb of a contig end. That conservative filter marked 1,544 of 9,694 BGCs (15.9%) as possible overestimates, including 41.5% of NRPS and 39.2% of PKS clusters.

Fragmentation mattered more for diversity than for raw counts. Excluding those weak fragments cut total BGCs by 16% but gene-cluster families (GCFs, the BiG-SCAPE grouping of related clusters) by 26%, from 4,669 to 3,452. Even after that filtering, seven metagenomes still produced 3,452 GCFs, a scale the authors compare with much larger global surveys while noting that sampling design and thresholds are not strictly equivalent across studies. Most clusters lacked close relatives in reference databases, which the paper reads as extensive novelty.

The product classes were dominated by ribosomally synthesized and post-translationally modified peptides (RiPPs, n = 1,361), terpenes (620), nonribosomal peptide synthetases (NRPS, 510), polyketide synthases (PKS, 367) and NRPS-PKS hybrids (107). GCFs were taxonomically widespread, led by Acidobacteriota (1,404) and Pseudomonadota (754), with 40 archaeal GCFs.

Only 9–27% of the encoded potential is transcribed

The central quantitative result is a gap between DNA and RNA: 9–27% of the recovered BGC potential was expressed in situ. This is a ratio of transcribed clusters to encoded clusters within this dataset, not a claim about BGCs in peatlands generally, and it rests on seven metagenomes from one fen. The authors describe the range as evidence that environmental and physiological constraints, rather than genomic capacity, govern secondary metabolism.

That gap is what makes the abundance-versus-activity comparison meaningful. Counting clusters tells you what chemistry a community could make; counting transcripts tells you which clusters were being used at sampling. The paper's own framing is that genomic capacity alone poorly predicts realized chemistry.

Acidobacteriota transcribe more while encoding less

Acidobacteriota accounted for over half of all BGC transcription despite only moderate repertoires. The contrast is with BGC-rich Pseudomonadota: taxa the paper calls talented producers, carrying up to 24 clusters, were largely silent. Across genomes, the fraction of a BGC repertoire that was expressed was inversely related to the number of BGCs per genome, quantified with Spearman's rank correlation.

The authors tie this decoupling to life-history strategy. Pseudomonadota had shorter predicted doubling times, estimated from genome-wide codon usage bias with gRodon2, and lower potential for complex carbon processing. Acidobacteriota were specialized for complex carbon processing and expressed a larger proportion of their repertoires. The interpretation offered is that large repertoires in fast-growing taxa represent latent capacity deployed only when regulatory and energetic conditions allow, conditions the authors argue are uncommon or short-lived in peatland. That is the authors' explanation of a correlation, not a demonstrated cause.

Expression tracks carbohydrate-active enzyme activity

At the genome level, the degree of BGC expression was most strongly coupled to carbohydrate-active enzyme (CAZyme) expression, particularly glycoside hydrolases (GHs), the enzymes that break down complex carbohydrates. Across all active MAGs, 97% of BGC-expressing observations also showed GH expression and 99.4% showed CAZyme expression.

The association held in models that accounted for genome and depth. GH expression was associated with BGC expression in a mixed logistic model (odds ratio 57.9, 95% CI 19.7–170). Greater GH expression was also associated with expressing a larger fraction of the encoded BGC repertoire after accounting for total MAG transcription, depth and MAG identity (odds ratio 4.78, 95% CI 2.73–8.35). A linear mixed-effects model with MAG identity and depth as random intercepts gave a population-level slope of β = 0.447 (95% CI 0.404–0.489) for normalized GH expression against BGC core-gene expression. Total CAZyme expression was nearly ubiquitous and therefore less discriminating as a binary predictor.

Redox context was heterogeneous by comparison. Among BGC-active focal MAG observations, expression co-occurred with microoxic or facultative terminal oxidases, aerobic oxidases, anaerobic-function markers, and low or unassigned redox modules. No single redox profile was shared across the 15 focal MAGs, which spanned Acidobacteriota, Pseudomonadota, Chloroflexota, Desulfobacterota, Desulfobacterota_B, Verrucomicrobiota and Actinomycetota.

Biosynthetic capacity declines with depth, expression does not stop

MAGs from deeper peat encoded fewer BGCs, but genome sizes were also smaller there, so the authors normalized by genome size. BGC density (BGCs per Mbp) still decreased with depth (Kruskal–Wallis H(4) = 16.2, P = 0.00275, η²H = 0.0216). The same direction appeared within Acidobacteriota (H(4) = 12.8, P = 0.0125, η²H = 0.0695). Genomes predicted as anaerobes harboured significantly fewer BGCs per genome than predicted aerobes (Welch's t-test, t(290.17) = 7.62, P = 3.56×10⁻¹³, Hedges' g = 0.68). MAG quality metrics did not differ across depths except for slightly higher completeness and GC content in deeper layers.

Expression, however, was detected at all sampled depths, including deeper anoxic layers, and occurred across a range of co-expressed redox contexts. The authors read this as secondary metabolism persisting where capacity is reduced and energy is limited, and as evidence that Acidobacteriota can express BGCs in both oxic surface and deeper anoxic zones.

What this fen supports and what it cannot settle

The evidence base is seven metagenomes from two cores in one fen, sampled in July 2020, with expression data missing for the deepest 20–30 cm interval and the surface 0–5 cm excluded. The 9–27% expression range and the dominance of Acidobacteriota are properties of this site and sampling window, not estimates for peatlands generally. The authors also note that cross-study comparisons of GCF richness are not strictly equivalent because sampling designs and analysis thresholds differ.

The mechanistic claims are correlational. The inverse relationship between BGCs per genome and expression fraction, and the coupling of BGC expression to GH expression, are associations quantified across MAGs and depths. The proposal that fast-growing Pseudomonadota hold latent repertoires awaiting rare permissive conditions is an interpretation consistent with the data, not a tested causal pathway.

The practical priority the paper establishes is a comparison, not a catalogue: within this fen, the taxa that transcribe the most BGCs are not the taxa that encode the most. Acidobacteriota, with moderate repertoires and active complex-carbon processing, contributed over half of BGC transcription, while cluster-rich producers stayed largely silent. Any survey that ranks biosynthetic importance by cluster count alone would invert that ordering here, which is the specific reason the authors argue that diversity cataloguing cannot substitute for measuring expression in situ.

Sources used in this answer