Grain feeding enriches gut bacteria that rapidly break down tryptophan

In the crowded, fermenting world of the cow rumen, a single microbial enzyme may hold the key to a long-standing problem in both animal health and human medicine: how to rein in a harmful metabolic product without destroying the beneficial microbes that produce it in healthy amounts. A new study published in the journal Microbiome by a research team at Nanjing Agricultural University reveals that when dairy cattle are switched to high-grain diets, their rumen microbial communities do not simply grow more indole by multiplying indole-producing bacteria. Instead, the ecosystem selectively enriches naturally occurring variants of the enzyme tryptophanase, or TnaA, that are intrinsically faster at churning out indole from the amino acid tryptophan. The finding reframes indole hyperproduction as an evolutionary ecological phenomenon rather than a simple overgrowth problem, and it points toward a new generation of precision inhibitors that can silence the pathological versions of an enzyme while leaving the beneficial ones untouched.

Indole is a deceptively simple molecule with a double-edged biological personality. At basal concentrations, it acts as a critical interkingdom signaling compound, promoting gastrointestinal homeostasis, reinforcing epithelial barrier function, and helping microbial communities coordinate their behavior. But when it accumulates at supraphysiological levels, indole becomes cytotoxic and pro-inflammatory, damaging host tissues and destabilizing the delicate microbial networks of the gut. In dairy cows, excessive ruminal indole is associated with the inflammatory and metabolic disturbances that accompany grain-heavy feeding, a management practice used to boost milk production but notorious for triggering rumen dysbiosis. Conventional interventions have relied on broad-spectrum antibiotics or nonspecific enzyme inhibitors, approaches that suppress pathological indole production but simultaneously wipe out the homeostatic signaling and commensal networks that depend on the same pathway. The therapeutic dilemma has always been precision: how do you distinguish pathological overproduction from healthy signaling when both come from the same enzyme?

The Nanjing team, led by Shengyong Mao and Hong Shen with co-first authors Jing Wang and Zhiming Xu, attacked the question at evolutionary scale. Rather than studying a handful of laboratory strains, they assembled a panoramic dataset of 462 TnaA orthologs drawn from 21,471 reference genomes and 197 rumen metagenomes, spanning the full breadth of sequence diversity in the microbial world. Sequences across this collection ranged from roughly 30 to 100 percent amino acid identity, a level of divergence that might suggest wildly different protein architectures. Yet when the researchers modeled each enzyme’s three-dimensional structure using computational approaches validated by metrics such as template modeling scores, a striking pattern emerged. Despite the extensive sequence divergence, every TnaA variant retained a highly conserved global fold, with structural similarity scores ranging from 0.89 to 1.00. The team describes this as a “conserved-fold, divergent-pocket” architecture: the overall protein scaffold is locked in place by evolutionary constraint, while the active-site pocket that binds tryptophan varies subtly but consequentially from variant to variant.

That structural variation translates directly into functional variation. Biochemical assays of representative enzymes revealed distinct active-site geometries and substrate affinities, meaning some TnaA variants convert tryptophan to indole far more rapidly than others. Phylogenetic reconstruction and genome-scale metabolic modeling both support the interpretation that this architecture reflects two hierarchically operating selective pressures within the bovine rumen. At the macro-evolutionary scale, the intense competition for colonization niches in the dense rumen environment imposes a purifying-selection-like constraint on the TnaA scaffold across diverse lineages, preserving the fold that makes the enzyme work at all. At the micro-evolutionary scale, however, diet-driven shifts in rumen fermentation change which kinetic variants thrive. Under low-grain conditions, the ecosystem favors attenuated TnaA variants whose modest indole output fits the molecule’s cooperative signaling role. Under high-grain conditions, the analysis found consistent enrichment of high-turnover variants, organisms carrying which gain a competitive advantage in the altered fermentation landscape, possibly because robust indole production helps them outmaneuver rivals in the crowded, rapidly shifting chemical environment of the grain-fed rumen.

The practical payoff of this structural insight came when the team went hunting for a molecule capable of exploiting it. Through screening and structural analysis, they identified luteolin, a naturally occurring plant flavonoid, as a variant-selective inhibitor of TnaA. The compound’s selectivity is remarkable: luteolin inhibits high-turnover TnaA variants with a half-maximal inhibitory concentration of 4.62 micromolar, while the sterically gated, attenuated commensal variants are roughly five and a half times more resistant, with an IC50 of 25.27 micromolar. Structural modeling explains the difference. In the high-turnover variants, the active-site pocket is unobstructed, allowing luteolin to penetrate freely and block the pyridoxal 5′-phosphate-dependent catalytic machinery. In the attenuated variants, subtle differences in pocket geometry create a steric gate that impedes the inhibitor’s access, sparing the enzyme’s function even at concentrations that fully disable the fast variants. In effect, the researchers discovered that evolution has already built a lock-and-key distinction between the pathological and homeostatic forms of the enzyme, and luteolin is the key that fits only one lock.

The technical achievement behind this discovery rests on a sophisticated integration of methods. The team combined evolutionary genomics, mining orthologs across thousands of genomes to reconstruct the enzyme’s phylogenetic history, with computational structure prediction using AlphaFold-style modeling assessed through predicted template modeling scores, local distance difference test metrics, and pairwise root-mean-square deviation calculations. Genome-scale metabolic modeling linked the presence of specific TnaA variants to predicted flux through indole-producing pathways under different dietary regimes, and biochemical validation with purified enzymes confirmed the kinetic predictions, including measurements of turnover against L-tryptophan substrate and inhibition curves for luteolin. Bayes empirical bayes analyses and likelihood ratio tests were used to detect sites under diversifying selection, while site concordance factors helped resolve the sometimes-tangled gene trees produced by horizontal gene transfer, a common phenomenon among rumen microbes that complicates any phylogenetic analysis.

Why does this matter beyond the dairy barn? The rumen serves here as a model ecosystem, a natural laboratory where diet manipulations reliably drive measurable microbial and metabolic shifts. The conceptual framework the authors articulate, that dysbiosis can manifest as enrichment of intrinsically hyperactive enzyme variants rather than proportional expansion of producing organisms, applies equally to the human gut, where indole hyperproduction has been implicated in inflammatory bowel disease, colorectal cancer progression, and other conditions. If pathological indole in the human intestine likewise arises from variant enrichment, then the strategy demonstrated here, targeting a structural vulnerability unique to hyperactive enzyme forms, could yield therapies that quiet the pathology while preserving the beneficial signaling that broad-spectrum approaches destroy. The work offers, in the authors’ words, a route from ecosystem-disrupting antimicrobials to function-selective microbiome modulation.

The study also carries immediate implications for livestock management. High-grain feeding is a cornerstone of modern dairy production, valued for its energy density and milk-yield benefits, but it is equally well known for triggering subacute ruminal acidosis, inflammation, liver abscesses, and reduced animal welfare. If the enrichment of high-turnover TnaA variants is a causal contributor to grain-induced ruminal dysbiosis, then luteolin or next-generation derivatives could be formulated as feed additives that specifically blunt pathological indole production during grain feeding, without disturbing the cooperative microbial networks that healthy rumen fermentation depends on. Because the sparing effect on attenuated variants is built into the structural biology rather than achieved through dosage compromise, the therapeutic window is unusually favorable compared with nonspecific inhibitors. The researchers caution, however, that their findings demonstrate association and mechanistic plausibility in model systems, and that controlled feeding trials will be needed to confirm efficacy and safety in living animals.

The evolutionary narrative embedded in the results is itself a contribution to microbiome science. The finding that a single enzyme family can be simultaneously conserved in fold and diversified in kinetics, with the diversification sculpted by diet, illustrates how quickly microbial ecosystems can shift their functional composition in response to environmental pressure. Horizontal gene transfer spreads TnaA genes across lineages, but selection on the enzyme’s kinetic properties determines which variants persist. This dual-scale view, macro-evolutionary constraint intertwined with micro-evolutionary ecological sorting, provides a template for studying other microbial metabolic functions implicated in dysbiosis, from short-chain fatty acid production to secondary bile acid metabolism and trimethylamine generation. The Nanjing group’s combination of large-scale comparative genomics with structural modeling and enzymatic validation offers a methodological roadmap others can follow.

For now, the image that lingers is one of remarkable molecular economy: a cow’s breakfast of grain cascades through rumen fermentation, shifts the chemical competitive landscape, and within that landscape, microbes carrying faster indole factories rise in abundance, their enzymes distinguished from their slower cousins by nothing more than a few subtle changes in the geometry of a catalytic pocket. That same pocket geometry, once mapped, becomes the target for a plant compound that can surgically disable the fast factory while the slow ones keep humming. It is a vivid demonstration that in complex microbial communities, the path to precise control runs through evolution’s own blueprint, and that the tools for treating dysbiosis may already exist in nature, waiting to be matched to the structural vulnerabilities that natural selection has left behind.

Wang, J., Xu, Z., Zhang, T. et al. Grain-induced rumen fermentation shifts drive ecological enrichment of high-turnover tryptophanase variants for precision microbiome modulation. Microbiome (2026). https://doi.org/10.1186/s40168-026-02508-4