The work, published in Frontiers in Microbiology, combines two complementary strategies: carefully tuned fermentation conditions that coax more product out of the cells, and targeted genetic rewiring of the biosynthetic machinery itself. Romidepsin, also known by its laboratory designation FK228, is a complex cyclic depsipeptide whose chemical synthesis remains impractical at scale, which makes microbial fermentation the only realistic production route. That constraint has kept the drug expensive and its supply chain fragile, so any genuine improvement in titer carries weight well beyond the laboratory bench.

Romidepsin’s biological activity stems from its ability to block histone deacetylases, enzymes that remove acetyl groups from histone proteins and thereby silence stretches of chromatin. By inhibiting these enzymes, the drug reopens tightly packed DNA, reactivates tumor suppressor genes, halts proliferation and pushes cancer cells into apoptosis. That mechanism earned it FDA approval, but producing it has been an upstream headache. The molecule is assembled by a hybrid trans-acyltransferase polyketide synthase and non-ribosomal peptide synthetase assembly line that stitches together cysteine, valine, dehydrothreonine derived from threonine, and malonyl-CoA through sequential condensation, macrocyclization and oxidative modification.

Earlier attempts to improve supply followed a winding path. Researchers identified the related thailandepsin gene cluster in Burkholderia thailandensis in 2011, and fermentation optimization of that system informed later work. Heterologous expression of the complete romidepsin gene cluster in Escherichia coli achieved only about 19 milligrams per liter, comparable to the native strain. Genome mining then revealed a highly homologous cluster in B. thailandensis MSMB43, which after minor modification produced 168.5 milligrams per liter. In 2022, the Shandong group took a decisive step: they deleted the native thailandepsin cluster from B. thailandensis E264 and replaced it with an attB integration site, creating a chassis strain called KOGC1 into which a hybrid romidepsin gene cluster could be slotted. That strain yielded 581 milligrams per liter, then the record, but still short of industrial requirements.

The new study began with physiology rather than genetics. The team grew the KOGC1 strain carrying an intact romidepsin gene cluster in M9 medium and systematically varied fermentation parameters using a sequential single-factor design. Romidepsin accumulation proved strictly growth-associated: titers climbed after 24 hours, peaked near 628 milligrams per liter at 60 hours and then declined, a pattern the authors attribute to end-product inhibition or nutrient depletion. A 60-hour harvest window was therefore fixed. Inoculum size mattered too; while 0.5 to 2 percent inocula performed similarly, a 5 percent inoculum depressed production, likely through high cell density and premature metabolite accumulation, so 1 percent was adopted.

The most striking fermentation lever was an adsorbent resin. Because romidepsin can be toxic to its own producer, the team added XAD-16 macroporous resin to the culture 12 hours after inoculation, letting the resin sequester the compound extracellularly as it formed. The authors speculate that this in-situ removal alleviates end-product toxicity and feedback inhibition, though they caution that direct evidence from intracellular and extracellular quantification, adsorption kinetics or cell viability assays was not obtained. Screening resin concentrations identified 4 percent weight per volume as optimal, roughly tripling the titer relative to resin-free controls. The team then supplemented the three amino acid precursors of the assembly line. Valine alone gave a marginal 6 percent gain, but feeding cysteine, valine and threonine together produced a synergistic effect, lifting the final titer to 743.5 milligrams per liter, a 28 percent increase over the 581 milligram baseline.

Genetic engineering delivered the second leap. Earlier work had shown that the AraC-family regulator TdpR acts as a positive activator of the cluster, and that including the tdpR gene boosted romidepsin yield 1.7-fold. The Shandong team reasoned that cranking up TdpR expression with strong promoters would push the whole pathway harder. Using RedEx technology, a method for seamless DNA insertion and deletion in large multimodular gene clusters, they inserted seven strong constitutive promoters upstream of tdpR. The promoters, drawn from a library of 50 natural promoters characterized in B. thailandensis E264 by luciferase reporter assays, were ranked by apparent strength from P46 down to P11.

The results contained a instructive surprise. Although the reporter assays had ranked P16 relatively low, in the actual chromosomal context P16 drove the highest romidepsin titer of all, reaching 1,109.4 milligrams per liter, a 90 percent increase over the previous KOGC1 baseline and a 49 percent gain over the fermentation-optimized intermediate. P46 and P33 also performed well, yielding 984.6 and 870.2 milligrams per liter respectively. Quantitative PCR confirmed that transcript levels of tdpR and two biosynthetic genes, depE and tdpJ, were elevated in the best-performing strains, with P16 showing the highest mRNA levels. Notably, the chromosomal transcription ranking did not match the reporter-based promoter strength ranking, a discrepancy the authors flag as evidence that surrogate reporter data poorly predict native in-locus transcription, a lesson for anyone designing promoter swaps in secondary metabolism.

The team also probed regulatory cross-talk. The romidepsin cluster dep and the thailandepsin cluster tdp are highly conserved in gene content and share a common assembly line, yet their dedicated regulators are strikingly divergent: TdpR is an AraC-type protein while DepR from C. violaceum is LysR-type, with no significant sequence homology. When the researchers cloned depR into the romidepsin expression vector and integrated it into KOGC1, the resulting strain maintained detectable romidepsin production comparable to the TdpR control. The authors suggest DepR may support the pathway through mechanisms distinct from TdpR rather than acting as a direct functional analog, though whether it binds the cluster’s promoters directly or works through an indirect regulatory cascade remains an open, testable hypothesis.

Perhaps as important as the headline titer is what did not happen over time. Large exogenous biosynthetic gene clusters often destabilize, deleting or rearranging under the metabolic burden they impose, which has blocked many natural product platforms from industrial translation. The team passaged three representative engineered strains for five generations without any antibiotic selection and found no significant titer fluctuation from the first through the fifth generation. Colony PCR on randomly picked colonies at generations one, three and five confirmed that the full refactored cluster remained intact, with no evidence of deletion, rearrangement or cassette loss. The authors note this demonstrates short-term genotypic and phenotypic stability only, and that bioreactor-scale and extended-culture performance remain to be assessed.

Honest caveats accompany the achievement. All experiments were conducted in 65-milliliter flask cultures, so the platform must be viewed as a foundation for future scale-up rather than an immediately industrial process. Late-stage broth viscosity, worsened by the resin, complicates downstream extraction, and the team suggests that optimized cell lysis protocols or continuous-flow separation systems will be needed to minimize product loss. Quantitative yield differences among engineered strains did not reach statistical significance under the current replication scheme, although the P16 strain’s titer was consistently reproduced across multiple independent fermentation batches. The authors also emphasize that direct metabolomic evidence for precursor limitation and product toxicity is still lacking, and they propose CRISPR-Cas-based genome stabilization and multi-omics profiling as next steps. Even with those qualifications, the study delivers the most efficient romidepsin production platform yet described and a generalized framework for stabilizing large, unwieldy gene clusters, work that could eventually make this potent anticancer agent cheaper, more abundant and more accessible to the patients who need it.

Subject of Research: Metabolic engineering of the FK228 (romidepsin) biosynthetic gene cluster for high-yield production in Burkholderia thailandensis

Article Title: Engineering the FK228 biosynthetic gene clusters for high-yield production

Article References: Wang, M., Jia, X., Zhang, Y., Ma, J., Zhang, Y., Fu, J., & Li, R. (2026). Engineering the FK228 biosynthetic gene clusters for high-yield production. Frontiers in Microbiology, 17, Article 1933794. https://doi.org/10.3389/fmicb.2026.1933794

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1933794

Keywords: FK228, romidepsin, HDAC inhibitor, Burkholderia thailandensis, biosynthetic gene cluster, promoter engineering, heterologous expression, XAD-16 resin, precursor feeding, TdpR, DepR, metabolic engineering