Modular biosynthesis of microbiome-derived polyketides and nonribosomal peptides: insights and opportunities for Lactobacillaceae

Abstract

Lactobacillaceae are a heterogeneous group of Gram-positive bacteria that play key roles in microbiome-dense ecosystems, particularly in mucosal habitats such as the vagina and gut, as well as in fermented foods. Widely recognized for their beneficial properties and broad applicability, they are best known for producing lactic acid and other small carboxylic acids. In contrast, their capacity to produce specialized metabolites, especially nonribosomal peptides and polyketides, has remained underexplored. These compounds, traditionally studied in soil-dwelling bacteria, are often assembled by large, multi-modular enzyme complexes, and include molecules with antibiotic, immunomodulatory and anticancer activities. Recent genomic surveys revealed that, despite their smaller genomes, Lactobacillaceae harbor numerous biosynthetic gene clusters encoding these metabolites, though only a handful has been experimentally characterized to date. Interestingly, the characterized examples display unusual domain architectures, atypical biosynthetic pathways, and novel chemical transformations. This review synthesizes current knowledge on the diversity, biosynthetic pathways and functional roles of specialized metabolites in Lactobacillaceae, and highlights their largely underexplored potential as one of the most important health-associated bacterial families known to date. Revealing their chemical repertoire not only expands our understanding of microbial specialized metabolism but also defines their roles in food preservation, plant health, and in preventing and treating infections across microbiome-rich mucosal ecosystems.

Background

Lactobacillaceae are a heterogeneous family of Gram-positive, (facultative) anaerobic, non-sporulating bacteria characterized by the production of lactate as the main end product of their carbohydrate metabolism [1]. Lactobacillaceae are core microbiome members of various environments, including plant surfaces, the gut of pollinators, the mucosal surfaces of mammals such as the human vagina and the chicken crop, as well as various fermented foods and beverages, where they can often dominate the ecosystem [1, 2]. In the human vaginal microbiome, for instance, Lactobacillaceae commonly form the dominant community whereas their depletion and the resulting increase in diversity have been associated with an increased risk of obtaining and transmitting bacterial and viral infections [3]. In vegetable fermentations, their dominance similarly reduces the likelihood of fungal and Enterobacteriaceae overgrowth while also providing health benefits upon consumption [4]. Even in niches where they do not dominate, Lactobacillaceae have still been linked to health. For example, taxa from this family are more prevalent in the healthy upper respiratory tract than in patients with chronic rhinosinusitis [5].

Since the beneficial effects of Lactobacillaceae on the gastrointestinal tract were first reported at the beginning of last century, their roles in human health, animal feed, and food processing and preservation have been extensively studied [6–10]. Much of this work has been focused on their role as probiotics, defined as “live microorganisms which when administered in adequate amounts confer a health benefit on the host” [11], and more recently on their development as live biotherapeutics, intended as medical products [12]. However, the exact mechanisms by which lactobacilli exert their beneficial effects and why they dominate specific environments and not others remain poorly understood [2].

The ability of Lactobacillaceae to produce large amounts of lactic acid and other small carboxylic acids, such as acetic acid and succinic acid, is often postulated as the main strategy that enables them to outperform other microbes and dominate specific environments and microbiomes [2]. Lactic acid production can rapidly lower the pH of the environment. Under these low pH conditions, lactic acid and other small carboxylic acids remain largely in their undissociated, membrane‑permeable form, allowing them to diffuse into sensitive cells. Once inside, they dissociate and acidify the cytosol, imposing stress on acid‑intolerant microbes and viruses [13, 14]. However, strains with similar lactic acid production and comparable pH‑lowering capacities can still exhibit markedly different antimicrobial activities, suggesting that additional mechanisms, including strain‑specific activities and specialized antimicrobial metabolites, contribute to their competitive success. Most research on these additional activities has focused on bacteriocins, a group of small antimicrobial ribosomally synthesized peptides [15, 16], while other specialized metabolites such as nonribosomal peptides (NRPs) and polyketides (PKs) have been largely overlooked in Lactobacillaceae.

NRPs and PKs are synthesized by large multi-enzyme complexes, called nonribosomal peptide synthetases (NRPSs) and polyketide synthases (PKSs), respectively, that act as molecular assembly lines [17, 18] (see Fig. 1). The use of a wide variety of building blocks and the installation of diverse post-assembly modifications creates extensive structural and functional diversity, resulting in a wide array of bioactive metabolites with significant pharmaceutical and industrial potential [17]. Examples from other microorganisms illustrate their broad applicability, e.g. as antibiotics, biosurfactants, parasiticides, herbicides, cytostatics, siderophores, immunosuppressives and anticancer agents [19–22]. In the context of the antibiotic crisis, NRPs and PKSs are of particular interest as they provide opportunities for discovering novel antibiotics with new or modified modes of action. In addition, the ‘generally-regarded-as-safe’ (FDA) [23] or ‘Qualified Presumption of Safety’ (EFSA) [24] status of many Lactobacillaceae makes strains that produce such antibiotics attractive candidates for live biotherapeutic applications, offering potential alternatives to traditional antibiotic treatments.

Fig. 1.

Simplified representation of the multi-modular structure of mega enzyme complexes working in a linear fashion. The common biosynthetic strategy is apparent. A Nonribosomal peptide synthetases uses (non-proteogenic) amino acid derivatives. B Polyketide synthases uses acyl-CoA derivatives (e.g. methylmalonyl, ethylmalonyl, allylmalonyl). Further explanation in text. Larger dark grey circles represent amino acid derivatives and pentagons represent acyl derivatives. A adenylation domain, C condensation domain, TE thioesterase domain, PCP peptidyl carrier protein domain (small light grey circles), AT acyltransferase domain, KS ketosynthase domain. Figure made with Inkscape

Although multiple NRPS and PKS gene clusters have been identified in Lactobacillaceae [25–30], only a small subset has been characterized in detail. The examples studied to date reveal unusual biosynthetic features, non-canonical domain architectures, and previously undescribed chemical structures and reactions. This review summarizes current knowledge on NRPS and/or PKS clusters characterized in Lactobacillaceae, with particular attention to their non-canonical features and to the structure and function of the resulting nonribosomal peptides (NRPs) and polyketides (PKs). By highlighting these unique biosynthetic pathways, we aim to illustrate the largely underexplored potential of Lactobacillaceae as a source of novel specialized metabolites.

The modularity of nonribosomal peptide synthetases and polyketide synthases

NRPS and PKS are often organized as large, multi-modular enzyme complexes or as stand-alone enzymes, which function in a sequential assembly-line manner. In most systems, each module is responsible for the incorporation and modification of a single building block [16, 27]. NRPS modules usually consist of three core domains (Fig. 1): a condensation (C), an adenylation (A), and a peptidyl carrier protein (PCP) domain (Fig. 1A). These domains work in concert to activate specific amino acid substrates, tether them to the enzyme complex, and catalyze peptide bond formation (Fig. 1A) [18]. PKS modules share an analogous architecture composed of a β-ketosynthase (KS), an acyltransferase (AT), and an acyl carrier protein (ACP) domain (see Fig. 1B), often supplemented with auxiliary domains, such as dehydratase (DH), ketoreductase (KR), and enoylreductase (ER) domains, that modify the β-keto-group of the ACP-tethered acyl thioester intermediates [31]. The strong organizational and catalytic similarity of NRPS and PKS systems enables the formation of hybrid NRPS/PKS biosynthetic gene clusters [32], which have also been identified in Lactobacillaceae, with the antimicrobial reutericyclin as a chemically characterized example [25].

Substrate selectivity in NRPSs is primarily controlled by the A domain with the active site consisting of 8–10 amino acid residues. These are referred to as the ‘specificity-conferring code’ [33, 34]. While NRPS systems collectively use a broad range of amino acids as monomers, over 200 characterized so far, including non-proteinogenic amino acids and α-hydroxy acids [35], predicting the monomer specificity of NRPS clusters in Lactobacillaceae remains challenging. This difficulty probably stems both from the strong bias in experimentally validated A‑domain reference datasets [36] toward well‑studied NRPS producers such as Streptomyces, Bacillus, Pseudomonas, and Burkholderia, and from the fact that many Lactobacillaceae NRPSs display non‑canonical domain architectures and atypical A‑domain motifs, which fall outside the scope of current prediction tools. Recently developed machine-learning-based predictors such as PARAS (and PARASECT), which are trained on expanded and more diverse A domain sequence-substrate datasets, may help overcome these limitations and improve prediction for systems from less-studied taxa [37]. In contrast, substrate specificity in PKSs is comparatively easier to predict. PKSs primarily rely on limited set of building blocks, such as malonyl-coenzyme A (CoA), methylmalonyl-CoA, ethylmalonyl-CoA, hydroxymalonyl-CoA, and methoxymalonyl-CoA, while starter units are more variable and may include more diverse CoA-thioesters, such as acetyl-CoA, propionyl-CoA, benzoyl-CoA, and various fatty acyl-CoAs [38, 39].

During chain elongation in both NRPS and PKS systems, activated monomers are covalently attached to the free thiol-group of the 4’-phosphopantetheine (4’PP) moiety of PCP (in NRPS) and ACP (in PKS) domains. This flexible 4’PP arm, derived from CoA, is linked to a conserved serine residue in the carrier protein by a phosphopantetheinyl transferase (PPTase) [40]. The C domain in NRPSs or KS domain in PKSs then catalyzes the elongation of the upstream growing chain with the newly incorporated monomer [18]. Product release is most often catalyzed by a terminal thioesterase (TE) domain [41]. TE domains use a water molecule or an internal nucleophilic group to catalyse the release of the fully assembled peptidyl/acyl chain as linear or a macrocyclic polypeptide or -ketide, respectively. Alternative release mechanisms exist, such as via thioester reductase (TR) and terminal condensation (C_T) domains, which release the compound as an aldehyde (or alcohol) or a macrocyclic peptide, respectively, as reviewed in detail by [42]. Further structural diversity arises from optional domains embedded that are within the enzymatic assembly line (acting in cis), free-standing (acting in trans) or acting following chain release, such as glycosyl transferases [43, 44], methyl transferases [45], oxidoreductases [46] and halogenases [47]. Due to the limited knowledge on PKS and NRPS clusters in Lactobacillaceae, examples of such optional domains have not yet been discovered.

The modular architecture and conserved active site motifs of PKS and NRPS enzymes make their corresponding genes well suited for genome mining approaches. Tools such as antiSMASH [48] and, historically, PRISM4 [49], have been central to predicting biosynthetic potential directly from genomic sequences. PRISM4, however, appears to be no longer maintained, as its webserver is currently unavailable, and no code updates have been released since 2020 (https://github.com/Adapsyn/prism-4-paper). In contrast, antiSMASH remains an actively developed platform, with biennial updates typically released alongside updates of the Minimum Information about a Biosynthetic Gene cluster (MIBiG) database [50]. The integration of complementary tools, such as the BiG-SCAPE/CORASON pipeline [51], has further enabled large-scale comparative analyses of biosynthetic gene cluster (BGC) diversity and facilitated connections with other omics datasets. More recent deep-learning genome mining tools, such as deepBGC [52] and GECCO [53], allow de novo detection of BGCs that may escape rule‑based method and they produce output formats compatible with antiSMASH. Despite this growing ecosystem of genome mining tools, antiSMASH remains the most widely adopted platform for specialized metabolite genome mining. However, in silico predictions provide only a first approximation: experimental validation remains essential to decipher the true biosynthetic output of BGCs, particularly in taxa such as Lactobacillaceae where non‑canonical architectures undermine prediction accuracy.

Distribution of NRPS and PKS types in Lactobacillaceae

NRPS and PKS systems are classified into different types based on their architecture and characteristics: (i) Type A NRPSs typically adhere strictly to the collinearity rule, with each module used only once in a defined sequence to incorporate a specific amino acid into the growing peptide chain; (ii) Type B iterative NRPS systems reuse modules multiple times to assemble their NRPs, whereas (iii) type C nonlinear NRPS systems deviate from modularity by reusing individual domains during biosynthesis [18]. Within Lactobacillaceae, the NRPS cluster of Lactiplantibacillus plantarum WCFS1 is predicted to be an example of a type A NRPS where each of the modules adds a single amino acid derivative to the final product [27]. To our knowledge, no type B or C systems have been characterized in Lactobacillaceae. Despite NRPSs being usually described as elaborate enzymatic assembly lines, stand-alone or dissociated nonmodular NRPS enzymes such as A/T didomains, single modules or free-standing domains have also been identified, for example by Patel, et al. (2022) in the tyrocitabine biosynthetic pathway expressed by L. iners LEAF2502A-d [26] as further discussed below.

PKSs are similarly subdivided based on catalytic domain organization: (i) type I noniterative, modular PKSs which typically follow the collinearity rule, (ii) type II iterative PKSs constituted by multiple trans-acting monofunctional stand-alone enzymes among which the active sites are distributed, and (iii) type III PKSs, which function independently of an acyl carrier protein [54]. Numerous variants exist within each category [55–57]. Both type I and II PKS systems are rare in Lactobacillaceae, with reutericyclin representing the only chemically characterized example produced by a hybrid, non-iterative trans-AT type I PKS/NRPS assembly line (Fig. 2) (see below). In a genome mining study by Rajput et al. (2023), 2,447 publicly available Lactobacillaceae genomes were screened and only 15 putative type I and 2 type II PKS BGCs were uncovered. Combined with other data, examples of type II PKS BGCs were predicted in Lapidilactobacillus mulanensis 143-6 [58], Limosilactobacillus oris [59], Ligilactobacillus salivarius An813 [58] and Limosilactobacillus reuteri isolates [29, 58] (see Table 1). The limited occurence of these types of PKSs in Lactobacillaceae suggests that this biosynthetic strategy is not widely conserved in Lactobacillaceae and is rather a strain-specific feature [58]. However, dedicated and targeted approaches are needed to functionally validate the relevance of these identified clusters. For Lactobacillaceae, but also in general, type III PKS systems remain poorly characterized compared to the other PKS types. Fortunately, recent updates to AntiSMASH (since version 6) have resulted in better detection of these type III PKS BGCs and compounds, which will likely result in an improved understanding [60]. Using the updated version 6, Rajput, et al. (2023) reported many putative type III PKS clusters in Lactobacillaceae members across all genera (except for Amylolactobacillus and Schleiferilactobacillus), ranging from highly adapted species, such as the insect-adapted Apilactobacillus and Bombilactobacillus, to those with free-living lifestyles (e.g. Lentilactobacillus and Levilactobacillus) [58]. Efforts to divide these clusters into cluster families resulted in over 65 different families which, in addition to our own screening (in preparation), suggests a quite conserved nature of these clusters particularly within species. Interestingly, no examples have been characterized to date, making their function still elusive, but showing a large, yet underexplored potential [58].

Fig. 2.

Overview of some well-studied NRPS/PKS clusters in LAB and their products. A Organizational structure of the best studied NRPS/PKS gene clusters in Lactobacillaceae. Color codes indicate the functional domains within the different genes. B Chemical structures. KS: ketosynthase, AT: acyltransferase, DH: dehydratase, KR: ketoreductase, C: condensation, A: adenylation, ACP: acyl carrier protein, PCP: peptidyl carrier protein, E: epimerization, MT: methyltransferase, TR thioreductase, TE thioesterase, PPTase phosphopantetheine transferase, NR nitroreductase. PCP/ACP domains are indicated as black boxes within the genes, the domain sizes are not indicative of the true size. *: prediction of the compound based on AntiSMASH (Version 8)

Table 1.

Overview of characterized and uncharacterized NRPS and PKS biosynthetic gene clusters in Lactobacillaceae from literature, omitting T3PKS due to their omnipresence

Species	Name or family	Year of discovery	Prediction	Level of documentation	Reference
Limosilactobacillus reuteri	Reutericyclin	2000	Hybrid	Knock-out	Höltzel et al. [62]; Rajput et al. [58]
Lactiplantibacillus plantarum	FAM_001	2003	NRPS	Genome mining	Kleerebezem et al. [27]; Elmasry et al. [93]; Rajput et al.[58]
Levilactobacillus brevis	FAM_002	2013	NRPS	Genome mining	Fukao et al. [28]; Rajput et al. [58]
Lactobacillus oris		2019	T2PKS	Genome mining	Sugimoto et al. [59]
Limosilactobacillus reuteri	fun	2019	T2PKS	Genome mining	Özçam, et al. [29]; Rajput et al. [58]
Limosilactobacillus reuteri	pks	2019	T2PKS	Genome mining	Özçam, et al. [29] Rajput et al. [58]
Limosilactobacillus fermentum		2020	NRPS	Metabolomics	Pavlova et al. [100]
Limosilactobacillus vaginalis	Reutericyclin-like	2020	Hybrid	Genome mining	Unpublished
Lactobacillus jensenii	FAM_003	2021	NRPS	Genome mining	Kumar and Dhanasekaran [101]
Lactobacillus iners	Tyrocitabine	2022	NRPS-like	Heterologous expression	Patel et al. [26]
Apilactobacillus quenuiae	FAM_004	2023	NRPS	Genome mining	Rajput et al. [58]
Lactiplantibacillus mudanjiangensis	FAM_005	2023	NRPS	Genome mining	Rajput et al. [58]
Lactiplantibacillus plantarum	FAM_002	2023	NRPS	Genome mining	Rajput et al. [58]
Lactiplantibacillus plantarum	FAM_006	2023	NRPS	Genome mining	Rajput et al. [58]
Lactiplantibacillus plantarum	FAM_007	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus apis	FAM_008	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus delbrueckii	FAM_009	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus helveticus	FAM_010	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus helveticus	FAM_011	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus helveticus	FAM_012	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus helveticus	FAM_013	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus helveticus	FAM_014	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus johnsonii	FAM_015	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus peromysci	FAM_016	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus psittaci	FAM_003	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus psittaci	FAM_017	2023	NRPS	Genome mining	Rajput et al. [58]
Lactobacillus taiwanensis	FAM_018	2023	NRPS	Genome mining	Rajput et al. [58]
Lapidilactobacillus concavus	FAM_019	2023	NRPS	Genome mining	Rajput et al. [58]
Lapidilactobacillus sp003946215	FAM_020	2023	T2PKS	Genome mining	Rajput et al. [58]
Leuconostoc gelidum	FAM_021	2023	NRPS	Genome mining	Rajput et al. [58]
Leuconostoc inhae	FAM_022	2023	hglE-KS	Genome mining	Rajput et al. [58]
Ligilactobacillus apodemi	FAM_023	2023	hglE-KS	Genome mining	Rajput et al. [58]
Ligilactobacillus salivarius	FAM_024	2023	T2PKS	Genome mining	Rajput et al. [58]
Limosilactobacillus reuteri	FAM_026	2023	NRPS	Genome mining	Rajput et al. [58]
Pediococcus pentosaceus	FAM_027	2023	NRPS	Genome mining	Rajput et al. [58]
Lapidilactobacillus achengensis	FAM_028	2023	NRPS	Genome mining	Rajput et al. [58]
Weissella hellenica_B	FAM_029	2023	NRPS	Genome mining	Rajput et al. [58]

Many putative NRPS/PKS systems have been predicted and reported in literature (Fig. 2; Table 1). Although most remain to be experimentally validated, those clusters that have already been characterized [26, 29, 61] show interesting chemistry, introducing interesting functional groups or catalyzing unusual reactions and are often of a non-canonical nature. In the following sections, we highlight a few key examples of systems identified and characterized in Lactobacillaceae.

Reutericyclin and its hybrid NRPS/trans-AT PKS assembly line

The first well-documented NRP/PKS-produced antimicrobial identified in Lactobacillaceae is reutericyclin A [25, 62]. Reutericyclin A comprises a tetramic acid (pyrrolidine-2,4-dione) ring system bound to a 2-decanoyl group (Fig. 3B). Its production was first described in Lm. reuteri LTH2584, initially isolated from an industrial sourdough, and this compound has since been detected in various strains of Lm. reuteri [62]. Homologous clusters have been identified in a few other lactic acid bacteria, such as Streptococcus [63] and Lactiplantibacillus [25]. Their presence in phylogenetically distant species suggests that horizontal gene transfer may have contributed to their dissemination, potentially facilitated by the cluster’s location within a horizontally acquired genomic island [25]. Notably, reutericyclin exhibits broad-spectrum activity against Gram-positive bacteria, including Bacillus subtilis, methicillin-resistant Staphylococcus aureus (MRSA), vancomycin-resistant Enterococcus faecium (VRE) and Clostridium difficile [64], leading to applications in therapeutics, skin care, and food fermentation.

Fig. 3.

The production of reutericyclin in L. reuteri LTH2584 and TMW1.656. A The genetic organization of the reutericyclin biosynthesis cluster. Color indicates the function of the gene, related to transport (green), regulation (blue), biosynthesis (light red) and Phosphopantetheinyl transferase (red). B The stepwise production mechanism of reutericyclin. The added monomers are indicated underneath each step and indicated in red in the chemical structure. Figure based on Gänzle et al. [61] and Lin et al. [25]

Although the chemical structure and antimicrobial activity of reutericyclin have been known since 2000 [61], the genes responsible for its biosynthesis were only identified in 2015 [25]. The reutericyclin gene cluster has been shown to typically contain nine coding sequences (CDSs), with five core biosynthetic genes (rtcABC and rtcNK) (Fig. 3A). The rtcNK genes encode a hybrid NRPS/trans-AT PKS assembly line, with RtcN containing a C, A and PCP domain, and RtcK featuring a KS-ACP-TE tridomain, activated by a putative trans-acting, unusual AT domain (RtcABC). The A domain of RtcN, while long believed to incorporate a D-leucine, appears to specifically select and activate an L-configured leucine residue as experimentally validated through [¹³C₁] L- and D-leucine feeding experiments [63]. As the final molecule contains the D-isomer of leucine, this suggests the involvement of a yet unknown epimerization reaction. The leucyl-PCP thioester undergoes condensation with 2-decanoyl-CoA in a reaction catalyzed by the C domain. In the second module, the abnormal trans-acting AT-like domain (RtcABC) activates malonyl-CoA and attaches it to the 4’PP-arm of the ACP domain (RtcK). The malonyl unit is then used to elongate the chain and the resulting product is released as a tetramic acid by the RtcK TE domain. Lin et al. (2015) proposed that RtcABC also adds an acetyl group at the 3-position on the tetramic acid ring, as RtcABC consists of three subunits that show sequence homology to the monoacetylphloroglucinol (MAPG) acyltransferase (PhlABC) [25]. This enzyme (PhlABC) is involved in the production of the inter- and intracellular messenger phloroglucinol from Pseudomonas sp. YGJ3 [65]. Nevertheless, the exact mechanism of RtcABC remains to be elucidated. The peripheral gene rtcP encodes a PPTase protein activating the peptidyl and acyl carrier protein domains. Deletion of either rtcRS or rtcT in Lm. reuteri TMW1.656 resulted in a lethal phenotype, suggesting roles in resistance and/or regulation. RtcR and RtcS are putative TetR family transcriptional regulators, affecting antimicrobial resistance and virulence; their deletion significantly reduced expression of RtcP. Additionally, truncation of the rtcN gene was shown to stop reutericyclin production, and reduced RtcP expression, indicating the existence of a positive-feedback mechanism. On the other hand, disruption of both rtcT and rtcRS in Lm. reuteri TMW1.656 increased sensitivity to reutericyclin, so that RtcT is predicted to be an MFS transporter exporting the active molecule [25].

The antimicrobial mode of action of reutericyclin appears similar to that of weak organic acids [64]. It selectively dissipates the transmembrane proton potential in sensitive cells, thus acting as a proton ionophore antibiotic. In food fermentation, reutericyclin can serve as a robust biological preservative agent. It is particularly effective against Bacillus subtilis, a spore-forming bacterium responsible for ropy spoilage in bread [66]. Since these spores survive baking temperatures, reutericyclin resistance to proteolytic and thermal degradation is essential for its activity. Lm. reuteri strains can produce reutericyclin up to physiologically relevant concentrations, i.e. a concentration high enough to inhibit Bacillus growth. Their addition during fermentation has been shown to extend bread shelf life [67]. Lin et al. (2024) further explored the ecological role of reutericyclin in food and intestinal ecosystems by comparing competitiveness between three isogenic Lm. reuteri strains: a wild-type reutericyclin producer, a reutericyclin-resistant non-producer and a sensitive non-producer. While reutericyclin had no impact on fitness during sourdough fermentation, the producer strain showed a clear in vivo ecological fitness advantage in the mice gut.

As mentioned, a wide range of Gram-positive bacteria are sensitive to reutericyclin, while Gram-negative bacteria, fungi, and yeasts appear resistant to reutericyclin. It has been shown that the hydrophilic nature of the lipopolysaccharide (LPS) structures in the Gram-negative outer cell membrane provides natural resistance to hydrophilic antimicrobials, like reutericyclin, since LPS mutant strains of Escherichia coli and Salmonella enterica became sensitive [62]. Notably, Gram-positive pathogenic MRSA and VRE strains, which are known for their multidrug resistance and difficulty to treat in current health care settings [68], are interesting targets for reutericyclin. However, due to the serum-binding properties of reutericyclin and cytotoxic activity [69], the applications of the molecule itself are limited to use as a disinfectant or topical application in a clinical setting. However, the use of their producer strain as therapeutic application, including live biotherapeutics, could be a viable alternative. This is of particular interest in a dermatological context as S. aureus is a frequent commensal skin colonizer that can cause severe infections and bacteremia [70], particularly when resistant strains like MRSA are involved. Reutericyclin can inhibit biofilm formation of S. aureus at levels comparable to ciprofloxacin [69], and is shown to eradicate S. aureus biofilms on superficial skin infections in mice. Its interesting activity suggests potential for skin products supplemented with either reutericyclin or reutericyclin-producing strains that could be used to efficiently treat skin infections. The supplementation of lactobacilli species and strains for the treatment of skin inflammatory lesions has already shown potential. For example, a probiotic topical cream containing micro-encapsulated live Lacticaseibacillus rhamnosus GG, Lactiplantibacillus plantarum WCFS1, and Lactiplantibacillus pentosus KCA1 bacteria showed a reduction of the relative abundance of staphylococci and Cutibacterium acnes, and reduction in inflammatory lesions in patients with mild-to-moderate acne symptoms both in an open label pilot study and in a double blind randomized placebo controlled trial with 80 patients [71]. However, to our knowledge, topical formulations containing reutericyclin or its producer strains have to date not been developed or made commercially available, possibly due to cytotoxicity [69, 72].

An antimicrobial and immune stimulatory polyketide

Other key examples of PKSs found in Lactobacillaceae are two predicted type II PKS systems in Lm. reuteri strains. Their discovery started with the observation that Lm. reuteri strains can activate the aryl hydrocarbon receptor (AhR), a key regulator of mucosal immunity [73, 74]. Upon binding, AhR translocates into the nucleus where it drives targeted gene expression, enhancing the innate immune system’s response and inducing the production of antimicrobial peptides by interleukin-22 signaling in the gut. This mechanism not only aids in the fight against intestinal pathogens but also protects tissues from inflammation damage.

By systematically assessing AhR activation across a wide range of Lm. reuteri strains, R2lc and 2010 were identified as the most potent activators [29]. Genome mining of these strains revealed the presence of T2PKS clusters. R2lc harbors two plasmids, i.e. pVP-R2lc01 and pVP-R2lc02, each with a distinct PKS-encoding gene cluster, named the fun and pks gene cluster, respectively. In contrast, strain 2010 was suggested to only harbor the pks gene cluster in the chromosome. Functional inactivation of the pks cluster of 2010 (53–87% amino acid identity to the pks cluster of R2lc) was linked to the observed AhR activation, further establishing this cluster at the basis for this strong activity [29], while the function of the fun cluster remains unknown.

Although the exact pathway for the pks-encoded compound remains unresolved, it is hypothesized to be similar to other type II PKS systems, showing some similarity to fatty acid synthesis. Based on the genetic content of the clusters, a putative ‘generalized’ pathway was proposed for both the fun and pks cluster (Fig. 4) [29]. In this model, malonyl-CoA is fed into the pathway by the multisubunit acetyl-CoA carboxylase (encoded by funH-K & pksK-N), then activated and loaded onto the ACP domain (encoded by funD and pksC/J). Subsequent condensation of two ACP-tethered malonyl units is catalyzed by the KS domain (encoded by funB and pksB). At this point in the pathway, there is likely a cycle where an additional malonyl unit is used to elongate the chain, with concomitant reduction and dehydration by the reductase-like and dehydratase domains encoded by pksDE/F and funE/L, respectively. The number of elongation rounds with malonyl-ACP building blocks is difficult to predict without a determined structure of the resulting compounds. After this cycle, the elongated product is likely further modified either by transamination catalyzed by an aminotransferase (encoded by funC) or by intramolecular addition by a metallohydrolase (pksG) potentially forming a lactone ring, similar to the ring in granadaene, shown in Fig. 2B. However, this modification is currently unknown: while the absorbance spectrum of the pigment could be determined [75], the full structure remains elusive (see Fig. 2B). By comparing R2lc and its non-producer R2lcΔpks, the chemical formula was determined as C₁₆H₁₆O₃ but to date no exact chemical structure has been determined [76]. The fun cluster shows homology to the granadaene pigment pathway from Group B Streptococci which produces a red pigment with hemolytic activity and as a result, the compound is predicted to be structurally related, but not identical to granadaene [77]. Notably, strains of Lm. reuteri have been shown to activate AhR by metabolizing dietary tryptophan to bioactive indole derivatives which trigger the AhR response whereas Lactobacillus johnsonii and Ligilactobacillus murinus did not [74]. However, this indole‑based mechanism has been demonstrated to not be the main drive of AhR activation for strains R2lc or 2010, likely due to the presence of the pks cluster in these isolates [73].

Fig. 4.

A The pks/fun biosynthetic gene clusters. KS ketosynthase, AT acyltransferase, DH dehydratase, KR ketoreductase, ACP acyl carrier protein, TE thioesterase, PPTase phosphopantetheine transferase. B Hypothetical pathway for the pks/fun compound. Inspired by Özçam, et al. [75]. Explanation in text

In addition to immunomodulatory activity, the pks compound was later found to function as a cell-cell contact-dependent antimicrobial [76, 78], giving the producer strains a competitive advantage over sensitive strains in intraspecies competition experiments in gnotobiotic mice. Comparative genomics and targeted gene deletions verified the role of an O-acyltransferase (encoded by act) in resistance to the compound [76]. This act gene expression resulted in an increased acylation of the cell wall, which is a common resistance mechanism against cell wall specific antimicrobials. While this suggests a cell wall-related mode of action, the precise mechanisms and interplay with the O-acyltransferase remains to be elucidated. Interestingly, this act gene seems to be widespread but not universally conserved in Lm. reuteri isolates, even those from diverse origins of isolation. Claus (2022) hypothesized that this acylation could improve bacterial fitness in the gut by protecting Lm. reuteri against host’s defense enzymes and antimicrobials. This serves as an example on how research on specialized metabolites cannot only identify novel compounds and potential live biotherapeutics but also broaden our understanding of ecological interactions.

Given their antimicrobial and immunological potential of the pks-encoded compound, Lm. reuteri strains are gaining interest to be applied as probiotic and live biotherapeutic for intestinal and vaginal health [79]. This is not surprising as Lm. reuteri exerts a multitude of activities as recently reviewed [79]. However, their antimicrobial spectrum should be phenotypically validated, and more (pre)clinical trials should be performed to further characterize and verify their applicatory potential. For instance, overactivation of AhR can theoretically be detrimental, instead of beneficial, leading to immune dysregulation and disease, so this should be carefully monitored [80]. Nevertheless, due to the prevalence of this species in a wide range of mucosal body sites in a variety of hosts, including humans but also livestock, potential applications are plentiful.

Tyrocitabines represent a novel class of antibiotics

The above-mentioned examples of specialized metabolite discovery have been mainly identified in Lm. reuteri isolates, a species known for its genetic accessibility, with tools available for genetic engineering and construction of gene deletion derivatives. Many other Lactobacillaceae taxa are more difficult to genetically access and engineer. This limitation forces many researchers towards the implementation of alternative strategies for characterizing specialized metabolite clusters such as heterologous expression. However, this is also not straightforward. One of the biggest drawbacks of heterologous expression, especially for multi-modular systems, is the large size of the genes and clusters and lack of good expression systems for Gram-positive organisms. In 2022, the group of Farren J. Isaacs introduced a cross‑kingdom heterologous expression platform that uses synthetic genetic elements to enable robust and efficient expression of putative antimicrobial gene clusters of unculturable or difficult-to-culture microbes from the human microbiome [26]. A key advantage of this system is that a single redesigned construct can be expressed across Gram-negative, Gram-positive, and fungal expression hosts. As a proof-of-concept, the authors expressed a highly unusual NRPS cluster from Lactobacillus iners, one of the most prevalent lactobacilli in the human vaginal microbiome. This cluster was cloned in E. coli and Pseudomonas putida, leading to the discovery of a novel antimicrobial class they named tyrocitabine. Tyrocitabine appears to inhibit on the post-transcriptional level, although its precise target is currently unknown. By systematically creating knock-out mutants lacking each of the predicted biosynthetic genes within the heterologous expression construct, they were also able to establish the exact synthesis pathway.

Within this cluster, the atypical role of the NRPS is noteworthy [26]. Unlike most NRPS systems, where the NRPS forms the core biosynthetic machinery [81], this is not the case in L. iners [26]. The biosynthetic pathway (Fig. 5) begins with the glycosylation of an L-tyrosine residue with phospho-ribosyl-diphosphate (PRPP) catalyzed by TybC, likely via an Amadori rearrangement, as characteristic intermediates were identified [25]. Next, TybE catalyzes a reduction reaction that favors the linear conformation, which is then followed by the addition of AMP to the free carboxylic acid group of the tyrosine (by TybB), leading to the formation of active tyrocitabine. Notably, TybB resembles a tyrosyl-tRNA synthetase, containing the signature Rossman fold but missing the tRNA binding domain. To this point, the NRPS has not been involved in the production of this compound. However, the NRPS (TybD) which consists of A-PCP-C domains, subsequently acylates the free hydroxyl groups of the AMP ribose residue, forming the inactive acylated prodrug. This modification likely increases stability and prevents self-toxicity of the producer strain. Upon export by TybA and contact with susceptible organisms, these acyl chains are presumably removed by extracellular esterases from microbiome members or extracellular oxidation, restoring antimicrobial activity.

Fig. 5.

A Tyrocitabine biosynthetic gene cluster. Color indicates the function of the gene, related to transport (green), biosynthesis (orange) and phosphopantetheinyl transferase (red). B Biosynthetic pathway as proposed by Patel et al. [26]

Due to the interesting activity of the TybB tRNA synthetase-like enzyme, a directed search with a Blast search binning clusters with at least two genes with at least 20% amino acid similarity to the proteins encoded within the tyb-encoded pathway of L. iners uncovered a diverse array of related clusters in many different clades of life, including Archaeal species and many bacterial Candidatus species. All identified clusters contained a TybB homolog, often in combination with NRPS, glycosyltransferase or other related genes, suggesting that tyrocitabine represents the first member of a broader class of specialized, nucleotide harboring metabolites [26].

Beyond its relevance for novel antibiotic discovery, this finding has important ecological implications for the vaginal microbiome. As mentioned in the introduction, the vaginal microbiome, compared to most other human-associated niches, is characterized by low species-level alpha diversity, typically dominated by one or a combination of Lactobacillus species, i.e. Lactobacillus crispatus, L. iners, Lactobacillus jensenii or Lactobacillus gasseri [82]. Depletion of these bacteria has been associated with an increased risk of obtaining and transmitting a plethora of infections. Therefore, their associations with vaginal health have been extensively studied. While the beneficial properties of L. crispatus are well established [83–86], the role of L. iners remains more controversial [87]. L. iners is found in both healthy and disease-associated states, such as bacterial vaginosis. The production of inerolysin, a pore-forming toxin produced by L. iners, similar to vaginolysin produced by the vaginal pathogen Gardnerella vaginalis, has been considered essential for its unwanted effects [87]. However, genomic studies increasingly suggest that L. iners comprises multiple subclades with distinct functional traits, indicating that virulence or beneficial properties are likely strain‑dependent [88, 89]. In this light, the discovery of tyrocitabine adds a new dimension to L. iners biology. Interestingly, 17% of vaginal human microbiome project samples of healthy women contained reads belonging to the tyrocitabine BGC, indicating an important contribution to the vaginal microbiome [90]. Yet, to our knowledge, these metabolites have not yet been detected in vaginal metabolomics samples. Patel et al. (2022) detected the tyrolose metabolite solely when examining its production in the native host L. iners LEAF 2052 A-d [26]. These findings imply that tyrocitabine might be reactive or unstable under laboratory conditions, consistent with observations reported for other microbiome‑derived metabolites [91, 92]. Importantly, the tyrocitabine (tyb) gene cluster is not universally present across L. iners strains and there is no evidence that its distribution correlates with the presence of inerolysin. This underscores the need for strain‑level analyses when assessing the ecological and clinical roles of L. iners. Given its potent antimicrobial activity, the role of tyrocitabine within the vaginal ecosystem warrants detailed investigation, and the purified molecule itself may hold therapeutic potential for preventing or treating infections and adverse reproductive outcomes such as infertility, preterm birth, and neonatal complications.

Uncharacterized multi-modular systems in Lactobacillaceae

Besides these well-characterized examples, there are many other gene clusters or metabolites that have been identified in Lactobacillaceae, but remain to be experimentally characterized. Especially in recent years, the improvement of mining tools such as antiSMASH [48] has led to the detection of previously undiscovered clusters. One prominent example is the plethora of Type III PKS clusters spread across the entire phylogenetic tree of the family of Lactobacillaceae [58]. Other clusters have been known for more than a decade but remain functionally unresolved despite multiple characterization attempts (including from our team), including the NRPS clusters of Lp. plantarum WCFS1 first mentioned in 2003 [27].

Although conserved across strains, no metabolite or biological function has yet been identified for the NRPS cluster of Lp. plantarum WCFS1. As yet mentioned above, Lp. plantarum WCFS1 is a well-studied strain, both in terms of its genome and its probiotic applications [27]. The NRPS cluster spans ~ 25 kbps and encodes 6 NRPS modules spread over two NRPSs, one with five modules and one with one module, consistent with type A NRPS architecture. Based on antiSMASH analysis (V8), we predict the production of a linear NRP (Fig. 2B), although it may also become circular. To verify this prediction, we obtained a Lp. Plantarum WCFS1 mutant with a deletion in the largest biosynthetic gene from the original team of NIZO food research (Ede, NL) [27]. However, comparative LC-MS analysis of the wild type and mutant revealed no differential peaks, suggesting that the compound may be expressed at low levels or only under specific conditions. This is a common challenge in specialized metabolite discovery, as many BGCs are conditionally expressed and remain silent under standard lab conditions. Attempts to identify triggers for improved expression proved difficult, and the large size of NRPS genes made heterologous expression technically demanding. As a result, the compound’s identity remains unresolved. In 2023, an antimicrobial polypeptide was isolated from another Lp. plantarum strain, LMG100, which was proposed to harbor an analogous NRPS cluster based solely on PCR detection with degenerate primers [93]. However, in the absence of a nonproducer mutant or further genetic validation, the evidence provided did not conclusively demonstrate that the polypeptide was of NRPS origin or that the activity of the purified supernatant was due to the NRP. Nevertheless, the similarity with a cluster from Streptococcus mutans with a role in biofilm formation and oxidative stress resistance [94, 95], suggests that the NRP of WCSF1 and LMG100 might exert similar functions.

Another example is the identified NRPS cluster in Levilactobacillus brevis KB290, first described in 2013 [28]. Lv. brevis is a species commonly identified in vegetable and cereal fermentations, is known as a spoilage organism for beer, and is also present in the human gut microbiome, making it an interesting potential probiotic [1, 96]. Known for its desirable antipathogenic traits [28], Lv. brevis KB290 has been used in fermented food products, such as suguki, a turnip variant. Despite its widespread use, the mechanisms behind its protective actions remain largely elusive. Sequencing of the strain KB290 genome in 2013 revealed nine plasmids, one (pKB290-3) containing an uncharacterized NRPS cluster [28]. Notably, Lv. brevis is known to adapt to certain environments by acquiring novel plasmids, maintaining a genetic diversity in its plasmidome [97]. It is tempting to speculate that this respective NRPS system could give the strain an advantage during food fermentation, but this remains to be substantiated. Of note, a similar cluster has been detected in a Lp. plantarum isolate [58]. Based on our antiSMASH analysis (v8), we predict that the KB290 cluster encompasses at least six genes, potentially ten, including an unusual, putative nitroreductase family gene. The NRPS itself consists of domains ordered as PCP-C-A-PCP, with the final product likely released by the downstream encoded TE domain. In close proximity to the NRPS cluster, a gene encoding a putative drug resistance transporter of the EmrB/QacA subfamily (LVISKB_P3-0008) was identified, suggesting a potential antimicrobial activity of the compound. To date, no additional studies have been published on this specific NRPS cluster, despite its intriguing genomic context and possible relevance for both probiotic functionality and antibiotic discovery.

Table 1 provides an overview of NRPS and PKS clusters or metabolites (with the exception for Type III PKS) from Lactobacillaceae isolates that have been mentioned in literature but lack functional, chemical, or in silico characterization. Recent advances in metagenomic and metatranscriptomic analysis highlight the underexplored potential. For instance, a recent systematic screening of over 30,000 lactic acid bacteria (LAB) genomes with AntiSMASH v6 has revealed over 1,000 NRPS and over 25,000 PKS clusters, besides 98.810 RiPP clusters [98]. While this broad analysis included all LAB taxa known to date, many were identified in genomes belonging to the Lactobacillaceae family, especially PKS clusters. Their relevance becomes even more apparent in the context of vaginal health. A recent meta‑analysis comparing 61 vaginal metagenomes from preterm births with 320 from full‑term pregnancies across three independent studies identified a total of 395 BGCs [99]. The authors reported that vaginal microbiomes containing a broader diversity of biosynthetic gene clusters, including NRPS, PKS and RiPP classes such as lanthipeptides and lassopeptides, was negatively associated with preterm birth [99]. However, this analysis remained high‑level. It did not resolve specific BGC families, nor did it distinguish clusters originating from Lactobacillaceae versus those from other, potentially pathogenic, taxa. Given the systematic screening data, it is plausible that at least some of the detected clusters were derived from vaginal Lactobacillaceae.

Conclusion

Lactobacillaceae possess a far richer and more diverse repertoire of NRPS and PKS biosynthetic systems than traditionally assumed, yet most remain uncharacterized at the functional or chemical level. To this regard, especially antimicrobial assays can be complicated by their ability to produce lactate and the resultant pH lowering, often necessitating pH neutralization, thereby risking that pH-dependent activities might be compromised. These isolates often reside in low pH environments, and their metabolites are likely optimal for these environments. Although targeted mutagenesis is the preferred strategy for genetically tractable species, many clusters remain silent under standard laboratory conditions, as illustrated by the NRPS system in Lp. plantarum WCFS1 [27], and genetic inaccessibility limits this approach for many other Lactobacillaceae taxa. Heterologous expression offers an alternative for clusters from difficult-to-culture isolates or metagenome-assembled genomes as shown for tyrocitabine from L. iners [26], and recent advances in synthetic biology demonstrate that even large, multimodular systems can be expressed across diverse hosts. Earlier assumptions that complex BGCs are rare in small genomes have proven incomplete, as Lactobacillaceae, including species with genomes below 1.5 Mbp [26, 58], harbor numerous multimodular systems with yet unknown ecological roles. Recent updates in genome mining tools, including antiSMASH v6 and later releases [48], have improved detection of BGCs, particularly T3PKS and combined with comparative genome mining [58, 59] now predicts hundreds of distinct cluster families within this group, suggesting that their biosynthetic potential has been substantially underestimated. Whether the discrepancy in biosynthetic diversity arises from insufficient database representation or genuine differences in biosynthetic capacity remains unclear. Future work combining large-scale comparative genome mining approach with functional characterization will be necessary to address these questions. Given their long history of safe use in food fermentation and widespread GRAS/QPS status, Lactobacillaceae metabolites and producers have potential for translational applications. Renewed exploration of these systems, combining large scale computational screening with targeted functional studies, is likely to uncover novel metabolites with relevance for microbial ecology, food biotechnology, and next generation probiotic and live biotherapeutic product development.

Abbreviations

4’PP

4’-phosphopantetheine

A

Adenylation domain

ACP

Acyl carrier protein

AT

Acyltransferase domain

BGC

Biosynthetic gene cluster

C

Condensation domain

CDS

Coding Sequences

C_T

Terminal condensation domain

CoA

Coenzyme A

DH

Dehydratase domain

E

Epimerization domain

ER

Enoylreductase domain

KR

Ketoreductase domain

KS

Ketosynthase domain

LAB

Lactic acid bacteria

MAG

Metagenome-associated genome

MIBiG

Minimum Information about a Biosynthetic Gene cluster

MT

Methyltransferase domain

NR

Nitroreductase domain

NRP

Nonribosomal peptide

NRPS

Nonribosomal peptide synthetase

PCP

Peptidyl carrier protein domain

PK

Polyketide

PKS

Polyketide synthase

PPTase

Phosphopantetheinyl transferase

RiPP

Ribosomally synthesized and post-translationally modified peptide

TE

Thioesterase

TR

Thioreductase

Author contributions

JD: Conceptualization, Investigation, Writing – original draft, review & editing, project administration and visualization. TE: Conceptualization, Investigation, Writing – original draft, review & editing, project administration. CD: Writing – review & editing, MVH: Writing – review & editing, DV: Writing – review & editing, JM: Writing – review & editing, SL: conceptualization, funding acquisition, resources, supervision, and writing – review and editing.

Funding

FWO research projects G049022N and G031222N; the industrial research fund UAntwerpen for IOF POC project CRUCIAL; and the European Research Council (ERC) for starting grant Lacto-Be (H2020) (grant ID 852600) and proof-of-concept VALERIE (Horizon) (grant ID 101213306). CD and MVH were funded by grants from Research Foundation – Flanders (FWO, grants 1S28622N and 1158725 N, respectively).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

S.L. declares to be a voluntary academic board member of the International Scientific Association on Probiotics and Prebiotics (ISAPP, www.isappscience.org), cofounder of YUN and scientific advisor for Freya Biosciences. She declares research funding from YUN, BioOrg, Puratos, DSM I-Health and Lesaffre/Gnosis. None of these organizations or companies were involved in the design, communication or data analysis. T.E. is partially funded by a VLAIO-project not related to this research. The other authors declare no competing interests related to this review manuscript.