Yeasts Lead the Race to Mass-Produce Next-Generation Antimicrobial Peptides

Antimicrobial resistance killed an estimated 1.14 million people worldwide in 2021, and resistance reservoirs now exist even for last-resort drugs such as colistin, a crisis driven in large part by heavy antibiotic use in livestock agriculture. A review published on 30 June 2026 in the open-access journal Fungal Biology and Biotechnology argues that fungi, and yeasts in particular, are the most promising manufacturing platforms for a new class of weapons against this threat: antimicrobial peptides, or AMPs. Written by Anna Donnan, Samuel M. M. Prudence and Sarah A. Kessans, the review surveys recent advances in fungal expression strategies and distils the engineering lessons that separate high-yielding production runs from expensive failures.

AMPs are small bioactive molecules, typically fewer than 100 amino acids and under 5 kilodaltons in mass, found across all three domains of life. Most are cationic and carry hydrophobic residues that let them fold into amphipathic shapes, allowing electrostatic interactions with microbial membranes that typically end in pore formation and membrane permeabilization. Because prokaryotic and eukaryotic membranes differ in composition, AMPs can selectively kill invading pathogens while sparing host cells, and in higher organisms they also play immunomodulatory roles. Their rapid bactericidal activity, broad specificity, high mobility and reduced likelihood of evolved resistance make them leading candidates to replace conventional antibiotics in medicine, agriculture and biotechnology. Modern tools have expanded the catalogue dramatically: as of 1 January 2026, the Antimicrobial Peptide Database documented more than 3379 natural, 2290 synthetic and 373 predicted AMPs, and site-directed mutagenesis now enables rationally designed libraries with enhanced activity, protease resistance, stability and target specificity.

The problem is making enough of them. Three production routes exist, and the review finds two of them wanting at scale. Direct purification from natural sources yields partially purified products in tiny quantities and can carry severe environmental consequences when the source species are rare or non-farmable; extracting the peptide apidaecin from the western honeybee requires roughly 50,000 bees to obtain a single gram of biologically active material. Chemical synthesis can deliver high purity for both natural and designed peptides, but its cost and time requirements are infeasible for commercial-scale output and become prohibitive for longer sequences or peptides with complex post-translational modifications. Heterologous expression, by contrast, is cheaper and scalable: while native extraction of apidaecin was estimated at 45 US dollars per gram, heterologous expression can produce it for about one dollar per gram.

Host selection then becomes the decisive variable. Bacteria grow fast on cheap feedstocks and have comprehensive expression toolkits, but they lack the cellular machinery for the eukaryotic post-translational modifications many AMPs require, and host toxicity frequently undermines bacterial production. Fungi, particularly yeasts, combine rapid growth and low nutritional requirements with the ability to perform essential PTMs and with strong secretion that simplifies purification for large-scale precision fermentation. Among yeasts, Pichia pastoris dominates. Its strong heterologous secretion capacity and its lack of ethanol production under aerobic conditions support the high cell densities that drive protein yields. The authors caution, however, that a popular engineering shortcut has delivered less than hoped: protease-deficient P. pastoris strains carrying deletions in genes such as pep4 and prb1 have been used to produce cecropin D, cecropin B, apidaecin and LL-37, but only two studies directly assessed the benefit of these deletions for AMP expression, and neither observed an increase in yield. With 104 predicted proteases across 50 families, Pichia’s proteolytic landscape remains largely unexplored.

Alternative yeasts show sporadic but instructive successes. Saccharomyces cerevisiae, the first yeast used for recombinant protein production, has underperformed for AMPs, partly because its metabolism diverts carbon to ethanol and limits biomass, and its hypermannosylated glycans can be immunogenic and hard to modify. In one study, expressing the O-glycosylated shrimp penaeidins Pen-2 and Pen-3a in S. cerevisiae, substitution of the native dimannosyl group with O-glycans significantly reduced activity in half of the tested strains, ruling the host out for those peptides. Saccharomyces boulardii, which survives gastrointestinal conditions better, was engineered to produce leucocin C active against Listeria monocytogenes, pairing production with in situ delivery. In a head-to-head comparison for two Enterococcus-derived AMPs, P. pastoris achieved a 21.4-fold increase in entA production over the native producer with 235-fold greater antimicrobial activity after 12 hours in YPD medium at 30 degrees, while Kluyveromyces lactis managed only a 1.2-fold production increase with reduced activity. Filamentous fungi such as Trichoderma reesei and Aspergillus species boast sophisticated secretion systems on paper, but few recombinant AMP reports exist, and none approach yeast yields, owing to larger genomes, complex fermentation, abundant AMP-destroying proteases and slower development timelines.

The review then dissects the levers that raise yield, starting with transcription. Most reported systems integrate the expression cassette into the genome rather than maintaining it on a plasmid, and targeted integration into transcriptionally active loci such as AOX1 avoids the off-target disruption and screening burden of random ectopic insertion. Gene copy number matters, but with a ceiling. P. pastoris X-33 transformants carrying one to fourteen copies of a protegrin-1 cassette showed linear yield gains up to six copies, while clones with more than eleven copies did not improve, and two six-copy clones expressed at different levels, hinting at additional controlling factors. Plectasin transformants with one, two, four and eight copies yielded 296, 409, 516 and 879 milligrams per litre respectively, and a four-copy Spgillcin177-189 integrant produced 126.1 milligrams per litre, a 2.75-fold gain over the single-copy strain, with no further improvement beyond that threshold. Metabolic burden, promoter saturation and secretion bottlenecks explain the diminishing returns. An elegant alternative is the 2A self-processing sequence, which exploits ribosomal skipping to release multiple discrete peptides from one mRNA: a four-copy 2A-based plectasin construct produced 183.2 milligrams per litre at 60.8 percent purity, a 2.3-fold increase, later upscaled to 426.3 milligrams per litre of total protein in a five-litre fermenter.

Promoter choice and codon design complete the transcription and translation picture. The inducible AOX1 promoter dominates the literature because it allows biomass to accumulate before peptide expression begins, and it delivered the highest reported titre in the review: 1505 milligrams per litre of NZL, a plectasin variant, in a five-litre fermenter. Inducer dosing is delicate, since insufficient methanol fails to drive transcription while excess is toxic to P. pastoris through formaldehyde and hydrogen peroxide accumulation. Constitutive promoters such as GAP are more practical industrially and can still hit high titres, with cecropin D reaching 485.24 milligrams per litre in a protease-deficient SMD1168 strain. Codon optimization is routinely applied but rarely tested as a controlled variable; in one of the few direct comparisons, codon-optimized mundticin ST4SA showed an eight-fold activity increase in S. cerevisiae while optimized plantaricin 423 showed no significant improvement, illustrating that the effects are peptide-specific and often masked by toxicity, proteolysis and secretion limits.

Post-translational engineering offers the richest toolkit. The yeast alpha-mating factor secretion signal has been used in nearly all P. pastoris AMP studies, but inefficient cleavage by the Kex2 and Ste13 proteases is a recurring failure mode, and in one study all nine alternative signal peptides tested for bovine Lfcin reduced production, most to undetectable levels. Hybrid signals that swapped only the pre-sequence solved the problem, with the 0030-alpha hybrid reaching 28.8 milligrams per litre and beating unhybridized MFα1. Peroxisomal targeting via an SKL signal peptide alleviated the host toxicity of LL-37 without loss of antimicrobial activity. Tags cut both ways: a SUMO fusion of the synthetic peptide L2 achieved 629 milligrams per litre while a FLAG fusion yielded roughly 50 percent less, a His-tag raised purification yield of sheep beta-defensin-1 from 35 to 80 milligrams per litre, yet a His-tag on corn defensin PDC1 reduced antifungal efficacy against Fusarium graminearum. Carrier proteins are rarer but striking: human serum albumin fused to apidaecin Ia yielded over 418 milligrams per litre of bioactive peptide, rice oleosin 18 fused to the antifungal PAF102 accumulated 180 milligrams per litre on lipid droplets with full activity retained, and the fluorescent protein iLOV enabled selection of high-expressing clones with up to 30 percent higher expression over a benchmark strain.

Fusing AMPs to themselves or to each other opens further doors. Tandem repeats of tachyplesin I released authentic monomers at 27.24 to 29.53 milligrams per litre with retained activity, while a tandemly arrayed plectasin multimer was secreted at 143 micrograms per millilitre with stability and activity comparable to the monomer, showing that cleavage is not always required. Hybrid dimers such as Lfcin-Lfa, expressed at 30 milligrams per litre, showed improved bactericidal activity over either subunit, and a linker-optimized chimeric construct combining porcine and avian cathelicidin variants with a solubility-enhancing four-helix bundle domain reached 79.02 milligrams per litre at 85.19 percent purity through a chromatography-free thermal precipitation process. Not every chimera works: a porcine beta-defensin-2 and interferon-gamma fusion showed no antibacterial activity, likely from improper folding. The most creative example repurposed proteolytic activation, replacing protegrin-1’s native elastase cleavage site with an enterokinase site so the peptide stayed inactive until reaching the duodenum, yielding 104 micrograms per millilitre and demonstrating intestinally targeted activation with reduced off-target effects. The authors close by calling for systematic head-to-head comparisons of codon strategies, copy number, promoters and strains, noting that current systems lean heavily on a narrow set of AOX1-based vectors and that fluorescence-based high-throughput screening could accelerate the search for robust, industrially scalable AMP production platforms.

Subject of Research: Fungal expression strategies for recombinant production of antimicrobial peptides

Article Title: Fungal expression strategies for heterologous production of antimicrobial peptides

Article References: Donnan, A., Prudence, S. M. M., & Kessans, S. A. (2026). Fungal expression strategies for heterologous production of antimicrobial peptides. Fungal Biology and Biotechnology, 13(1), Article 12. https://doi.org/10.1186/s40694-026-00220-4

Image Credits: AI Generated

DOI: 10.1186/s40694-026-00220-4

Keywords: antimicrobial peptides, antimicrobial resistance, Pichia pastoris, heterologous expression, yeast, filamentous fungi, codon optimization, promoter engineering, protein fusions, secretion signals, gene copy number, peptide dimers