Romidepsin, an approved treatment for T-cell lymphomas, belongs to a family of bicyclic depsipeptides that inhibit class I histone deacetylases, HDACs. Every member of this family shares a conserved Zn2+-binding pharmacophore, yet each carries a structurally distinct peptidyl cap that modulates isoform selectivity by contacting the outer rim of the HDAC active-site tunnel. Biosynthetically, the pharmacophore and cap are assembled by separate polyketide synthase, PKS, and nonribosomal peptide synthetase, NRPS, machineries that must interface precisely. How the conserved pharmacophore apparatus engages productively with structurally diverse cap biosynthetic systems, enabling the natural combinatorial biosynthesis that generates this structural diversity, was largely unresolved.
Researchers in the Gregory L. Challis and Lona M. Alkhalaf Groups at the University of Warwick and the ARC Centre of Excellence for Innovations in Peptide and Protein Science at Monash University, published in Nature Communications, used gene proximity searching of the antiSMASH database to locate the long-sought biosynthetic gene cluster, BGC, for FR-901375 in Pseudomonas chlororaphis subsp. piscium DSM 21509. The team then combined in vitro reconstitution of chain elongation across the PKS-NRPS interface, site-directed mutagenesis, AlphaFold 2 structural modelling, carbene footprinting mass spectrometry, and molecular dynamics simulations to dissect the docking mechanism that links the conserved pharmacophore machinery to the variable cap machinery across four related assembly lines.
The FR-901375 BGC in P. chlororaphis subsp. piscium DSM 21509 encodes a four-module NRPS, PcdK, predicted to assemble a D-Val–D-Val–D-Cys–L-Thr tetrapeptidyl cap. Gene deletion of pcdK abolished FR-901375 production, confirming BGC identity. The docking elements at the PKS-NRPS junction, a short linear motif, SLiM, appended to the C-terminus of the final pharmacophore ACP domain and a β-hairpin docking, βHD, domain at the N-terminus of the first cap NRPS subunit, proved to have asymmetric contributions. In-frame deletion of the βHD domain eliminated FR-901375 production in vivo and abolished condensation activity in vitro, whereas deletion of the SLiM caused only a modest reduction comparable to mutation of a single hydrophobic residue to a charged isostere, V101D, in the ACP domain. Comparative BGC analysis suggests the FR-901375 BGC evolved from a spiruchostatin BGC ancestor via horizontal transfer of a burkholdac NRPS gene followed by successive duplication and recombination events that remodeled cap assembly while retaining pharmacophore machinery.
AlphaFold models of the ACP-SLiM/βHD-C di-domain complex revealed that the βHD domain engages not only the SLiM, as seen in the enacyloxin IIa and other previously studied systems, but also binds directly to a conserved epitope on the globular ACP domain. A hydrophobic protuberance formed by Val101 and Thr102 at the C-terminus of the last ACP α-helix nestles into a hydrophobic pocket on the βHD domain, while Arg21 and Glu97 on the ACP domain form salt bridges with each other and with residues Asp12 and Glu97 of the βHD domain. Carbene footprinting mass spectrometry confirmed masking at the predicted ACP-βHD interface and at the ACP-C domain contact region, consistent with the modelled complex. Mutagenesis of Arg21, Glu97, Val101, and Thr102 each reduced condensation activity, validating the interaction epitope. Productive in vitro crosstalk between all pairwise combinations of ACP-SLiM di-domains and βHD-C-A-PCP tetra-domains from the FR-901375, romidepsin, burkholdac, and spiruchostatin systems confirmed that the conserved ACP-βHD epitope enables non-cognate pharmacophore and cap machineries to engage functionally.
These findings establish the molecular basis for combinatorial biosynthesis within the depsipeptide HDAC inhibitor family and trace a plausible evolutionary path for BGC diversification. The conserved ACP-βHD interaction epitope identified here provides a structural framework for engineering hybrid PKS-NRPS assembly lines, with potential applications in generating new analogues of romidepsin and related anticancer agents with altered cap structures and isoform selectivity profiles. The work also offers a rational starting point for identifying the still-elusive BGC for largazole by defining the bioinformatic signatures shared across this compound class.
