Ornithine lipids and other acyloxyacyl amino lipids: the coming-of-age story of a group of non-canonical membrane lipids

Abstract

Abstract

Ornithine lipids (OLs) are phosphorus-free membrane lipids present in many bacteria, but absent from eukaryotes and archaea. Three pathways for OL synthesis have been reported to date. Conditions that induce OL synthesis include elevated temperature, low pH, low phosphate concentration, and low salt concentration. OLs can be modified by different hydroxylations, N-methylation, or taurine transfer. These modifications can be expected to alter the biophysical properties of individual lipid molecules and the membrane as a whole, with potential applications in synthetic biology. The presence and synthesis of OLs are frequently associated with increased stress resistance, and bacterial mutants of some species deficient in OL synthesis show increased susceptibility to elevated temperatures or reduced pH. OLs have been shown to be important for bacteria-host interactions and, recently, to interact with Toll-like receptor 4 (TLR4). We present a comprehensive analysis of the taxonomic distribution of genes encoding putative OL synthases, enabling predictions of which bacteria are expected to have the capacity to synthesize OL at least under specific growth conditions. Lipids structurally analogous to OLs in which other amino acids replace ornithine have also been described and are synthesized by enzymes homologous to OL synthases. In recent years, a wide range of studies and observations related to OLs have been published, including the identification of genes encoding novel OL synthases, novel OL-modifying enzymes, and novel OL structures; the sensing of OLs and other aminolipids by eukaryotic organisms; and their possible use in synthetic biology. In the present review, we discuss these recent advances.

Key points

• Ornithine lipids are phosphorus-free membrane lipids present in a wide range of bacteria.

• The presence and induction of OLs are associated with increased stress resistance.

• The presence and modification of OLs affect the membrane properties of E. coli cells.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00253-026-13777-2.

Introduction

All cells are surrounded by membranes that serve as a permeability barrier. These membranes are composed of amphipathic lipids, which have both hydrophobic and hydrophilic portions. In aqueous environments, these amphipathic lipids tend to self-assemble: hydrophobic portions associate, and hydrophilic portions interact with each other and the aqueous phase, forming micelles or bilayer structures. In bacteria, the best-known and best-studied membrane lipids are glycerophospholipids, such as phosphatidylglycerol, phosphatidylethanolamine, and cardiolipin. These and other glycerophospholipids are found in all studied bacteria, including the model organism Escherichia coli (Raetz and Dowhan 1990). For several years, it was thought that the bacterial membrane lipid composition was of low complexity and static. We now know that the composition of bacterial membranes can be highly diverse, reflecting the taxonomic diversity of bacteria, and can even change within the same organism in response to changing environmental and growth conditions (Sohlenkamp and Geiger 2016). Other glycerophospholipids, such as phosphatidylcholine, phosphatidylserine, and phosphatidylinositol, have been identified in certain bacteria, and, in addition to this complexity, several more lipid classes have been described. Some membrane lipids present a diacylglycerol backbone structure, such as betaine lipids, glycosylated diacylglycerols, and sulfolipids like sulfoquinovosyl diacylglycerol (SQDG), whereas others, for example, hopanoid and steroid lipids, sphingolipids, sulfonolipids, and ornithine lipids (OLs), lack this structure (Sohlenkamp and Geiger 2016). Many of these lipids are formed only by certain bacterial groups or under specific growth conditions.

OLs belong to a larger group of lipids called acyloxyacyl aminolipids, which differ in the amino acid(s) forming the hydrophilic part of the amphiphilic lipid (Liu et al. 2024). The α-amino group of the headgroup amino acid is N-acylated by a 3-hydroxy fatty acid, and a secondary fatty acid is esterified to this 3-hydroxy group. Such an acyloxyacyl structure, sometimes called piggyback structure, is also present in lipid A (Raetz et al. 2007). Ornithine is the amino acid most frequently described as the headgroup of aminolipids, but lysine, glutamine, glycine, the dipeptide serine-glycine, and the tripeptide ornithine-serine-glycine are present in some taxonomic groups (Bill et al. 2021; Moore et al. 2016; Sohlenkamp 2017). These aminolipids have been detected only in bacteria and are absent from archaea and eukaryotes. Based on genome analysis, approximately 50% of sequenced bacterial species were estimated to harbor genes encoding OL synthases (Vences-Guzmán et al. 2015).

Some bacterial species form OLs constitutively, whereas in others, their synthesis can be induced. Several authors have observed that the presence of OLs increases under or is restricted to specific stress conditions (Barbosa et al. 2018; Córdoba-Castro et al. 2020; Geiger et al. 2010, 1999; González-Silva et al. 2011; Keck et al. 2011; Palacios-Chaves et al. 2011; Pitta et al. 1989; Sandoval-Calderón et al. 2015; Vences-Guzmán et al. 2011, 2025).

In recent years, a wide range of studies and observations related to OLs have been published. These include the identification of genes encoding novel OL synthases, novel OL-modifying enzymes, and novel OL structures; the mechanisms by which OLs and other aminolipids are sensed by eukaryotic organisms; and their possible use in synthetic biology to construct membranes with desired properties for biotechnological applications. The purpose of the present review is to give an overview and discuss these recent advances.

Methodology

Taxonomic distribution of OL synthases

To evaluate the taxonomic distribution of OL synthases across bacteria, query protein sequences for OlsB (accessions Q92SJ1 from Rhizobium meliloti, Q9HW51 from Pseudomonas aeruginosa, Q7D1N1 from Agrobacterium fabrum, and Q9RCY9 from Streptomyces coelicolor), OlsF (accessions A0A7U0N5K6 from Serratia proteamaculans and A5FLQ1 from Flavobacterium johnsoniae), OlsH (accession L0DAW6 from Singulisphaera acidiphila), and GlsB (accession Q8A247 from Bacteroides thetaiotaomicron) were retrieved from UniProtKB (UniProt 2025). The enzymatic activities of all these queries have been demonstrated experimentally. A reference protein database was constructed from complete, annotated, and reference bacterial genomes from RefSeq (taxon 2) (Goldfarb et al. 2025), downloaded in August 2025 using NCBI Datasets (v.18.5.1) (O'Leary et al. 2024) (Supplementary datasets S1), along with their complete taxonomic information. A local BLAST database was created from these RefSeq proteins using makeblastdb, and homology searches were performed with BLASTp (v.2.16.0+) (Camacho et al. 2009) using each query sequence against the local database with an E-value threshold of 1e−10 and a maximum of 10,000 target sequences per search. BLASTp results were filtered based on stringent coverage and similarity criteria, applying thresholds of ≥ 30% identity, ≥ 70% query coverage, and ≥ 70% subject coverage. All hits passing these criteria were retained, allowing multiple hits per species (Supplementary datasets S2, S3, S4, S5). From the filtered results, taxids of homologous sequences were extracted and cross-referenced with RefSeq taxonomic information to generate taxonomic distributions by phylum, class, and order for each enzyme (Supplementary data S6, S7, S8, S9).

To construct a representative phylogenetic species tree reflecting bacterial diversity, one taxon was randomly selected from each taxonomic order present in RefSeq taxonomy data previously downloaded (Supplementary Dataset S10). Phylogenetic marker identification was performed using HMMsearch (v.3.4) (hmmer.org) with Hidden Markov Model profiles from Pfam (PF00562, PF04563, PF04997, PF04998, and PF11987), corresponding to conserved RNA polymerase subunits, as described (Witwinowski et al. 2022). Searches were conducted against the selected taxa using gathering thresholds (--cut_ga). Sequences of the three phylogenetic markers were independently aligned using MAFFT (v.7.525) (Katoh and Standley 2013) with the L-INS-i algorithm, and alignments were refined with trimAl (v.1.5.rev0) (Capella-Gutiérrez et al. 2009) using the automated method. The three trimmed alignments were concatenated, and the phylogenetic tree was constructed using IQ-TREE (v.3.0.1) (Wong et al. 2025) with 1000 ultrafast bootstrap replicates and default parameters. Final trees were visualized using iTOL (Letunic and Bork 2021) with custom annotation files that included taxonomic labels and binary datasets indicating the presence/absence of OlsB, OlsF, and OlsH enzymes in each taxonomic order represented in the tree. For the annotation of presence or absence, any taxonomic order containing at least one homologue was marked as present, regardless of the total number of homologues identified.

Synthesis and diversity of OL structures

Although OLs have been known to be formed by some bacteria for decades, their study has been difficult until, in the early 2000 s, the OlsBA pathway for OL synthesis was first described in the α-proteobacterium Sinorhizobium meliloti (Gao et al. 2004; Weissenmayer et al. 2002). At that time, far fewer genome sequences were available, but genes encoding OlsBA homologues were quickly described in the genera Rhodobacter, Burkholderia, and Pseudomonas, belonging to the α-, β-, and γ-proteobacteria, respectively. In the OlsBA pathway, the first step is catalyzed by the N-acyltransferase OlsB that transfers a 3-hydroxy fatty acid from the constitutive acyl carrier protein AcpP to the α-amino group of ornithine, leading to the formation of lyso-OL (LOL) (Fig. 1, reaction 1). In the second step, the O-acyltransferase OlsA transfers a fatty acid from AcpP to the 3-hydroxy group of the first fatty acid, leading to the formation of the unmodified OL (Fig. 1, reaction 2). Interestingly, OLs had also been detected in diverse bacterial species whose genomes had been sequenced but lacked genes encoding OlsBA, examples being Serratia marcescens, Desulfovibrio sp., Flavobacterium sp., and Sorangium cellulosum (Kates et al. 1964; Keck et al. 2011; Makula and Finnerty 1975; Pitta et al. 1989). Motivated by this observation, Vences-Guzmán and coworkers (Vences-Guzmán et al. 2015) identified the bifunctional OL synthase OlsF in Serratia proteamaculans. OlsF harbors an O-acyltransferase domain and an N-acyltransferase domain fused together (Fig. 1, reaction 3). The first step, the N-acyltransferase reaction, is catalyzed by the C-terminal domain of OlsF, and the second step, the O-acyltransferase reaction, is catalyzed by the N-terminal domain of the protein. The fatty acid donors for the OlsF-catalyzed reactions have not been determined. The discovery of OlsF as a second OL synthesis pathway in addition to OlsBA also enabled an updated prediction of the capacity for OL synthesis based on the analysis of bacterial genome sequences. Genes encoding OlsF homologues were detected in some bacterial genera known to form OLs, but that lack OlsBA homologues. Interestingly, genes encoding OlsF homologues were also present in some well-studied organisms, such as Vibrio cholerae and Klebsiella pneumoniae, where their presence had not been reported (Barbosa et al. 2018; Olea-Ozuna et al. 2025; Vences-Guzmán et al. 2015, 2025).

Fig. 1.

Synthesis of OLs and their derivatives. Shown are the three known biosynthetic routes to the unmodified OL shown in the center of the figure. In some cases, OLs may undergo multiple modifications, leading to structures not shown in this figure. The names of the OLs, S1 (substrate 1) and S2 (substrate 2), originally described their roles as substrates in the OlsC-dependent OL modification reaction. In an analogous manner, P1 (product 1) refers to its role as a product in the OlsC-dependent OL modification reaction. The lipid NL1 (new lipid) was identified as an unknown lipid in B. cenocepacia. The colored background of the figure facilitates distinguishing OL synthesis (upper part, reactions 1–5) from OL modification (lower part, reactions 6–10). In the center of the figure, the unmodified OL S1 is shown. OL modifications introduced by the modifying enzymes are highlighted in different colors. AcpP, constitutive acyl carrier protein; α-KG, α-ketoglutarate; SAM, S-adenosylmethionine; SAHC, S-adenosylhomocysteine

When studying the N-methylation of OLs in the planctomycete Singulisphaera acidiphila, it was observed that its genome lacked genes encoding OlsBA or OlsF, suggesting the presence of a third pathway for OL synthesis (Escobedo-Hinojosa et al. 2015). Recently, OlsH, an enzyme responsible for lyso-OL formation in S. acidiphila and related to plant GH3 proteins (Jez 2022), was identified and characterized (Fig. 1, reaction 4) (Rivera-Najera et al. 2025). The OlsH-catalyzed reaction uses ATP and a fatty acid as substrate with AMP and pyrophosphate as leaving groups, similar to the long-chain fatty acyl-CoA synthetase FadD (Black et al. 1992), but in the case of OlsH, no acyl-CoA is formed, and a lyso-OL is formed directly. The second step for OL synthesis in this pathway, the O-acyltransferase reaction, was also identified recently and is catalyzed in S. acidiphila by the α/β-hydrolase OlsI (Fig. 1, reaction 5) (Rivera-Najera et al. 2025). This pathway appears to be conserved, at least in several planctomycetes, but OlsHI homologues are widespread also among bacteria of the Fibrobacterota, Chlorobiota, and Bacteroidota (FCB) superphylum (Fig. 2).

Fig. 2.

Bacterial species tree with distribution of predicted OL synthases. Bacterial representative phylogeny based on the concatenation of RNA polymerase subunits β, β′, and elongation factor IF-2, rooted between the clades of Terrabacteria and Gracilicutes (as in Witwinowski et al. 2022). The phylogenetic tree was constructed using IQ-TREE with 1000 ultrafast bootstrap replicates. The innermost track shows the number of proteomes used per taxon, displayed as a gradient color strip from light purple (low proteome count) to dark purple (high proteome count). The middle track displays the binary presence/absence pattern of each key enzyme involved in the synthesis of ornithine lipids (OL): OlsB (red), OlsF (orange), and OlsH (green). The outermost track shows the relative abundance (percentage) of each protein per order

The existence of three independent OL synthesis pathways is interesting because it implies that OL synthesis pathways have evolved at least three times convergently. OLs are more frequent than other aminolipids having different amino acids in their headgroup, and homologues of the enzymes involved in the synthesis of OLs have been shown to be responsible for the synthesis of aminolipids with other amino acid headgroups in some cases (Liu et al. 2024; Lynch et al. 2019; Olea-Ozuna et al. 2025; Smith et al. 2019; Vences-Guzmán et al. 2015).

Although the unmodified OL appears to be most common (OL, S1, at the center of Fig. 1), various OL modifications have been reported in a range of bacteria. Five different OL modifications have been described so far: OL can be hydroxylated at three different positions, the δ-amino group of the ornithine headgroup can be N-methylated, and taurine can be transferred to the carboxy group of the ornithine headgroup (Fig. 1, reactions 6 to 10). More than one modification can be present in the same OL molecule, and the introduction of these modifications potentially alters membrane properties.

The 2-hydroxylation of the ester-bound fatty acid had already been described several years ago in different bacterial species (Fig. 1, OL P1), such as Gluconobacter cerinus, Burkholderia cenocepacia, and Flavobacterium sp., and appears to be the most widespread OL modification (Vences-Guzmán et al. 2012). This specific hydroxylation has long been recognized in various species of the genus Burkholderia, to the point that it has been considered a taxonomic marker (Córdoba-Castro et al. 2020; Taylor et al. 1998; Yabuuchi et al. 1992). The gene olsC, encoding the OL hydroxylase OlsC, was described during a search for genes involved in the response to acid stress in Rhizobium tropici (Rojas-Jiménez et al. 2005), and later it was shown that OlsC introduces a 2-hydroxylation in the secondary fatty acid (Fig. 1, reaction 6) (Vences-Guzmán et al. 2011). The gene/enzyme responsible for introducing a hydroxylation at the same position in OL in B. cenocepacia and Flavobacterium sp. has yet to be identified.

Looking in B. cenocepacia for the gene/enzyme responsible for introducing the 2-hydroxylation in the secondary fatty acid, specifically for an OlsC homologue, a gene encoding a 2-hydroxylase of the primary fatty acid, olsD, was described (Fig. 1, reaction 7, OL NL1) (González-Silva et al. 2011). OlsC and OlsD are homologues of LpxO from Salmonella, which has been shown to introduce a 2-hydroxylation into a lipid A-bound secondary fatty acid (Gibbons et al. 2000, 2008). All three enzymes belong to the same family of Fe²⁺/O₂/α-ketoglutarate-dependent oxygenases. In addition to Burkholderia genomes, genes encoding OlsD homologues can be found in the genomes of Serratia sp. and Mesorhizobium sp. (Diercks et al. 2015; Vences-Guzmán et al. 2015). Consistently, it has been shown that Mesorhizobium loti presents an OL hydroxylated at the 2-position of the amide-bound 3-hydroxy fatty acid (Fig. 1, OL NL1) (Diercks et al. 2015).

A gene encoding a hydroxylase that introduces a hydroxylation in the C4-position of the ornithine headgroup was discovered in R. tropici (Fig. 1, reaction 8, OL S2) (Hölzl et al. 2018; Vences-Guzmán et al. 2011). OlsE belongs to the di-iron fatty acyl hydroxylase superfamily (cl01132), which includes fatty acid and carotene hydroxylases as well as sterol desaturases. Apart from R. tropici, OlsE activity has been detected in the plant pathogen Agrobacterium tumefaciens (Vences-Guzmán et al. 2013).

Another modification of OLs that was unknown prior to 2013 is the N-methylation of the δ-amino group (Fig. 1, TMOL). This modification was described in some planctomycetes (Moore et al. 2013), and motivated by this finding, Escobedo-Hinojosa et al. set out to identify the OL N-methyltransferase OlsG (Fig. 1, reaction 9). OlsG is related to phospholipid N-methyltransferases and, like the latter, it can use phosphatidylethanolamine as substrate, causing the formation of phosphatidylcholine upon heterologous expression in E. coli (Escobedo-Hinojosa et al. 2015).

Interestingly, the hydroxylations of the primary fatty acid and the ornithine headgroup (González-Silva et al. 2011; Hölzl et al. 2018; Vences-Guzmán et al. 2011), and the N-methylation (Escobedo-Hinojosa et al. 2015; Moore et al. 2013), were unknown until relatively recently, showing the power of mass spectrometry for lipid analysis and discovery (Moore et al. 2016). There is clearly a need for continued MS-lipid analysis throughout bacterial species.

For several decades, the presence of a taurine-modified OL, cerilipin, has been known in Gluconacetobacter cerinus (Fig. 1, reaction 10). Using enzyme assays, the enzyme has been characterized to some extent, and it was established that taurine transfer to hydroxylated OL is ATP-dependent (Tahara et al. 1976a, 1978). Nevertheless, its identity and the encoding gene remain unknown.

It is unclear whether additional OL modifications exist and what the functional implications of the OL modifications described so far might be with respect to membrane properties. One possible advantage of OL modifications is that they enable rapid responses and adaptation of the bacterial membrane to changing environmental conditions without requiring de novo lipid synthesis. It is clear that we need to continue studying OLs in greater detail to understand their functions.

Distribution of genes encoding OL synthases in bacteria

OLs were first described more than 60 years ago in mycobacteria and Rhodopseudomonas spheroides (Gorchein 1964; Laneelle et al. 1963). Over the following two decades, they were detected in more bacteria, but few bacterial species seemed able to form them, so the general impression was that OLs were uncommon and exotic lipids compared to glycerophospholipids. From the 1990 s onwards, many bacterial genomes were sequenced, and after Otto Geiger and coworkers had discovered the genes encoding the OlsBA pathway in S. meliloti (Gao et al. 2004; Weissenmayer et al. 2002), it became possible to learn about the capacity of organisms to synthesize OLs by searching their genomes for the presence of genes encoding putative OL synthases. Later, the bifunctional OL synthase OlsF was first identified in S. proteamaculans, and recently, the OlsHI pathway was described in S. acidiphila (Rivera-Najera et al. 2025; Vences-Guzmán et al. 2015), indicating that at least three distinct OL synthase pathways exist. We searched for the corresponding homologues in bacterial genomes to obtain an updated estimate of the extent of the capacity to form OLs. Protein sequences of key enzymes involved in OL synthesis were used as queries, and BLASTp results were filtered using stringent coverage and similarity criteria, with thresholds of ≥ 30% identity, ≥ 70% query coverage, and ≥ 70% subject coverage. We used OlsB as an indicator of the OlsBA pathway and OlsH as an indicator of the OlsHI pathway. For all queries used in the search, their enzyme activity had been confirmed. The presence of genes encoding OL synthases in bacterial genomes was mapped onto a bacterial species tree (Fig. 2). For the presence/absence annotation, any taxonomic order containing at least one homologue was marked as present, regardless of the total number of homologues identified. The tree also includes the number of proteomes per order in the analysis and the relative abundance of each homologue within each order, showing the actual abundance of these homologues in the bacterial tree of life.

OlsB was first discovered in S. meliloti, a member of the family Rhizobiaceae, which belongs to the order Hyphomicrobiales. Genes encoding OlsB homologues are present in several other orders of α-proteobacteria, such as Geminicoccales, Rhodospirillales, Acetobacterales, Sneathiellales, Emcibacterales, Kordiimonadales, Sphingomonadales, Hyphomicrobiales, and Rhodobacterales, but are absent from the orders Rickettsiales, Ca.-Pelagibacterales, Ca.-Puniceispirillales, Parvularculales, Hyphomonadales, Maricaulales, and Caulobacterales. Within the β-proteobacteria, genes encoding OlsB homologues are found in the orders Rhodocyclales, Neisseriales, Burkholderiales, Ferrovales, and Nitrosomonadales. Within the γ-proteobacteria, genes encoding OlsB homologues are present in the orders Cardiobacteriales, Nevskiales, Chromatiales, Methylococcales, Moraxellales, Pseudomonadales, Oceanospirillales, and Alteromonadales (Fig. 2). OlsB activity and genes encoding functional OlsB homologues have been described in diverse genera of the α-, β-, and γ-proteobacteria classes.

Genes encoding OlsB homologues were detected in several species of the orders Kitasatosporales, Streptosporangiales, Glycomycetales, Micromonosporales, Sporichthyales, Nakamurellales, Mycobacteriales, Actinopolysporales, and Pseudonocardiales from the phylum Actinomycetota. For Streptomyces coelicolor from the order Kitasatosporales and Mycobacterium tuberculosis from the order Mycobacteriales, the presence of OLs has been demonstrated, and for S. coelicolor, a mutant lacking the olsB gene does not form OLs (Sandoval-Calderón et al. 2015).

Genes encoding OlsB homologues have also been detected in some bacterial orders with no prior reports of OL. Within the phylum Cyanobacteriota, genes encoding OlsB homologues were relatively abundant within the order Nostocales. It has been reported that some cyanobacteria remodel their membranes under phosphate limitation, with increased synthesis of the phosphorus-free sulfolipid SQDG, but OLs have not been identified in these organisms (Caille et al. 2024). Similarly, genes encoding OlsB homologues are present in the case of a few species of the orders Bacilliales and Eubacteriales from the phylum Bacillota (previously called firmicutes), where the presence of OLs or other aminolipids has not been described. Within the phylum Spirochaetota, in the order Leptospirales, genes encoding OlsB homologues can be detected in a major proportion than in the previous case, although again, we could not identify reports of the presence of OL in the literature.

OlsF was first described in the γ-proteobacterium Serratia proteomaculans, which belongs to the order Enterobacterales (Vences-Guzmán et al. 2015). Within the γ-proteobacteria, genes encoding OlsF homologues can also be identified in the orders Thiohalomonadales, Lysobacterales, Chromatiales, Methylococcales, Pseudomonadales, Oceanospirillales, Cellvibrionales, Kangiellales, Aeromonadales, Immundisolibacterales, Alteromonadales, and Vibrionales. Vibrio cholerae is unique in that its genome encodes two functional OlsF homologues, whose expression is induced by two distinct stress conditions. The expression of one homologue is induced under conditions of low phosphate in the growth medium, whereas the expression of the second homologue is induced by reduced salt concentrations (Vences-Guzmán et al. 2025).

Within the phylum Pseudomonadota, genes encoding OlsF homologues can also be found in the orders Magnetococcales, Emcibacterales, Burkholderiales, and Nitrosomonadales. Genes encoding OlsF homologues are frequent in the phylum Thermodesulfobacteriota, which includes the group of bacteria previously called δ-proteobacteria, and are identified in the orders Desulfuromonadales, Desulfobacterales, Desulfobulbales, Desulfovibrionales, Desulfarculales, Desulfobaccales, and Syntrophobacterales. The presence of OLs in members of this bacterial group is well known (Seidel et al. 2013; Sohlenkamp and Geiger 2016). Genes encoding OlsF homologues are also present in the phylum Campylobacterota, within the orders Nautiliales and Campylobacterales, which had been classified as ε-proteobacteria previously. A functional OlsF homologue causing the formation of OLs and lysine lipids (LLs) when expressed heterologously in E. coli has been identified in Flavobacterium johnsoniae (Vences-Guzmán et al. 2015), belonging to the order Flavobacteriales, and genes encoding OlsF homologues can also be detected in the orders Cytophagales, Sphingobacteriales, Bacteroidales, and Marinilabiliales of the phylum Bacteroidota.

Finally, genes encoding OlsF homologues can be detected in orders within the Planctomycetota, Verrucomicrobiota, Chlamydiota (PVC) superphylum, such as Lentisphaerales, Tepidisphaerales, Verrucomicrobiales, and Puniceicoccales, and in other orders as Fidelibacterales, Ignavibacteriales, Balneolales, Bryobacterales, and Terriglobales, but without functional evidence of OlsF activity or OL presence. The lack of functional evidence for many OlsB and OlsF homologues may reflect that some of the orders in which genes encoding OL synthases are present have been studied little. Moreover, because much of the lipidomics performed to date has been targeted, previous studies may not have sought OLs, even in otherwise well-characterized organisms.

The OlsHI pathway for OL synthesis has been identified so far only in the planctomycete Singulisphaera acidiphila (Rivera-Najera et al. 2025). OlsH catalyzes the formation of an N-acyl bond between the α-amino group of ornithine and the carboxyl group of a fatty acid that is activated with ATP, and we think it is possible, similarly to OlsB and OlsF, that OlsH homologues outside the planctomycetes might use a different amino acid as substrate. N-acyl amino acids are known to be produced by bacteria from diverse taxonomic groups, and it is often unclear how they are synthesized. In S. acidiphila, a member of the order Isosphaerales, OlsH has been shown to catalyze the formation of N-acyl ornithine. Genes encoding OlsH homologues are also present in planctomycetes from the order Gemmatales. Outside the planctomycetes, genes encoding OlsH homologues are present in bacteria from the orders Opitutales, Rhodothermales, Cytophagales, Saprospirales, Chitinophagales, Sphingobacteriales, Flavobacteriales, Bacteroidales, Marinilabiliales, Acanthopleuribacterales, Geobacterales, Rhodospirillales, Acetobacterales, Rhodobacterales, Acidithiobacillales, and Vibrionales.

In summary, genes encoding putative OL synthesis are present in many bacterial orders, but they are far more common in diderm bacteria than in monoderm bacteria.

Functions of OLs

Some bacteria form OLs constitutively, whereas others only form them under specific growth conditions. Among the bacteria that have been described to form OLs constitutively are Brucella sp., Burkholderia sp., Agrobacterium sp., Mesorhizobium loti, and R. tropici (Córdoba-Castro et al. 2020; Devers et al. 2011; González-Silva et al. 2011; Palacios-Chaves et al. 2011; Rojas-Jiménez et al. 2005; Sohlenkamp and Geiger 2016; Taylor et al. 1998; Vences-Guzmán et al. 2013, 2011). For many other bacteria, it has been described that OLs accumulate only under specific stress conditions. Among the abiotic stresses known to induce OL formation are the limitation of phosphate in the growth medium (Gao et al. 2004; Geiger et al. 1999; Weissenmayer et al. 2002) and low salt concentration (Vences-Guzmán et al. 2025). In addition, the presence of OLs has been shown to increase bacterial resistance to low pH and elevated temperatures (Bedoya-Pérez et al. 2024; Vences-Guzmán et al. 2011).

Often, examples being Pseudomonas aeruginosa, S. meliloti, Vibrio cholerae, and S. proteamaculans, it has been observed that no OLs (or only minor amounts) are formed when bacteria are cultivated in complex growth media commonly used as standard laboratory media. These media usually contain excess phosphate and other nutrients, not necessarily reflecting the conditions found in the natural habitats of the bacteria. When cultivating these bacteria in defined growth media with a drastically reduced phosphate concentration, a large fraction of the glycerophospholipids can be substituted with phosphorus-free membrane lipids such as OLs, DGTS, or SQDG (Barbosa et al. 2018; Benning et al. 1995; Geiger et al. 1999; Lewenza et al. 2011; Minnikin and Abdolrahimzadeh 1974; Minnikin et al. 1972, 1974; Vences-Guzmán et al. 2015, 2025).

It has been speculated that the zwitterionic OL functionally replaces the zwitterionic PE, mainly based on the reciprocal amounts observed in the membrane under different growth conditions and their shared zwitterionic properties (Geiger et al. 2010; López-Lara et al. 2005, 2003). Biophysical studies, however, revealed important differences between the two lipids, suggesting that the prevailing view may be overly simplistic (Mukhina et al. 2022). Using differential scanning calorimetry, X-ray scattering, and X-ray fluorescence of OL in mono- and bilayers, it was observed that OL showed a greater tendency than chain-analogous PE to form ordered structures, including a molecular superlattice based on a hydrogen-bonding network between the headgroups. These results indicated that OL and PE behave very differently in ordered single-component membranes, but obviously, this does not exclude the possibility that OLs may behave more similarly to PE in fluid multicomponent membranes (Mukhina et al. 2022). This membrane remodeling occurring under conditions of phosphate concentrations is regulated by the transcriptional regulator PhoB in S. meliloti, and in S. proteamaculans, the olsF gene is also preceded by a pho box (Geiger et al. 1999; Vences-Guzmán et al. 2015). Interestingly, in bacteria that form OLs constitutively, a decreased phosphate concentration in the growth medium appears to increase OL content (Vences-Guzmán et al. 2011).

It has been suggested that this remodeling allows the liberation of phosphate bound within membrane glycerophospholipids and its use in other cellular processes, such as nucleic acid formation. This hypothesis is consistent with the findings published by Bedoya et al. (2024), which demonstrated that a synthetic E. coli strain forming OLs and accumulating reduced amounts of PE grows to higher cell densities at a limiting phosphate concentration than a wild-type E. coli strain, which only forms the phospholipids PE, phosphatidylglycerol, and cardiolipin. This strain forming OLs produces more biomass than the wild-type in a growth medium with the same phosphate concentration, which could have useful applications in industry.

In nature, most habitats are probably oligotrophic, whereas the laboratory situation in which microorganisms are grown under nutrient excess is highly artificial. Therefore, the capacity to remodel the membrane should be an advantage for bacteria with diverse lifestyles (Zavaleta-Pastor et al. 2010). Remodeling could improve bacterial survival in phosphate-depleted habitats, enabling adaptation. Often undetectable under phosphate-replete conditions, phosphorus-free membrane lipids form far more than half of the total membrane lipids under phosphate-limiting conditions (Geiger et al. 1999; López-Lara et al. 2005; Vences-Guzmán et al. 2015, 2025). Following this logic, it might be expected that mutants deficient in OL formation show a clear growth phenotype, but this is not always the case. S. meliloti mutants deficient in OL formation grow as the wild-type under phosphate-limiting conditions. To observe a growth phenotype under these conditions, OL deficiency must be accompanied by DGTS deficiency (López-Lara et al. 2005). Likewise, a S. proteamaculans mutant deficient in OlsF, which cannot synthesize OL, grows similarly to the wild-type under phosphate-limiting conditions (Vences-Guzmán et al. 2015). In contrast, V. cholerae mutants deficient in the OlsF homologue VC0489, which is induced under phosphate-limiting conditions and that cannot form OL under these conditions, exhibit reduced growth (Vences-Guzmán et al. 2025).

Recently, it was described that the V. cholerae genome encodes two different OlsF homologues, VC0498 and VCA0646 (Fig. 1, reaction 3). Both were shown to be functional when expressed in E. coli. VC0489 was shown to be induced by phosphate limitation, whereas lower salt concentrations induced the transcription of VCA0646 and the synthesis of OLs (Barbosa et al. 2018; Vences-Guzmán et al. 2025). Several members of the genus Vibrio are halophiles, a characteristic that limits their ability to colonize particular niches. Pandemic V. cholerae strains, however, can grow in a relatively broad range of salinity concentrations, a characteristic that allows them to survive in brackish water and freshwater (Baker-Austin et al. 2018). Changes in salinity might require adaptations in the cell membrane properties, which could involve the formation of OLs.

The presence of hydroxylated OLs was shown to be important for stress tolerance and resistance, including increased temperatures and low pH, and exposure of bacteria to these stresses can increase the accumulation of certain OL species. Although no increase in the formation of hydroxylated OLs can be observed when R. tropici is cultivated at 37 or 42 °C instead of 30 °C, a R. tropici mutant deficient in the OL 2-hydroxylase OlsC (Fig. 1, reaction 6) grows much more slowly than the wild type at 42 °C. Taylor et al. (1998) observed an increase in hydroxylated membrane lipids (including hydroxylated OL and hydroxylated PE) in Burkholderia cepacia grown at 40 °C, supporting the idea that OL hydroxylation has a function at higher temperatures. Similarly, as described for hydroxylated sphingolipids and hydroxylated lipid A, the presence of an additional hydroxyl group might allow for the formation of additional hydrogen bonds between the hydrophilic parts of the membrane lipids and contribute to membrane stabilization (Nikaido 2003).

In different bacteria, OLs have been shown to be enriched in the outer membrane (Dees and Shively 1982; Palacios-Chaves et al. 2011; Vences-Guzmán et al. 2011), and interestingly, genes encoding putative OL synthases are more abundant in diderm bacteria than in monoderm bacteria (Fig. 2). For Thiobacillus thiooxidans, it has been suggested that the presence of OLs in the outer membrane might be related to its acid tolerance (Dees and Shively 1982). R. tropici accumulates increased amounts of 2-hydroxylated OL P1 (Fig. 1) when grown at pH 4.0, and R. tropici mutants deficient in OlsC (Fig. 1, reaction 6) grew much more slowly than the wild-type under these conditions (Vences-Guzmán et al. 2011). Consistent with the idea that the additional hydrogen bonds might contribute to the stabilization of the membrane (Nikaido 2003), Bedoya et al. (2024) showed a shortening of the lag phase when a synthetic E. coli strain forming unmodified OL and OlsC-hydroxylated OL (Fig. 1, OL P1) was exposed to mildly acidic conditions.

A few decades ago, it was shown that phosphate limitation increased the resistance of Pseudomonas fluorescens to polymyxin B, and this resistance was correlated to the presence of OLs (Dorrer and Teuber 1977). Later, in P. aeruginosa, it was shown that although polymyxin resistance increased under phosphate-limited conditions, this phenotype was apparently not caused by OL accumulation (Lewenza et al. 2011). A second study, also using P. aeruginosa as a model, subsequently reported a contradictory finding: OL accumulation increased resistance to polymyxin B and altered cell-surface properties (Kim et al. 2018). Recently, two independent studies reported an apparent correlation between the presence of OLs and increased bacterial tolerance to cationic peptides in two other bacterial species (Olea-Ozuna et al. 2025; Vences-Guzmán et al. 2025). Acinetobacter baumannii does not form OL when grown in complex media, but does so under phosphate limitation. Under phosphate-limiting conditions, OL accumulates, and the bacteria show increased resistance to colistin (also called polymyxin E) (Olea-Ozuna et al. 2025). V. cholerae grown under low salt conditions, which induced the OL synthase VCA0646 and OL accumulation, were more resistant to cationic peptides than the mutant deficient in VCA0646 grown under the same conditions (Vences-Guzmán et al. 2025). As OLs also accumulate in the outer membrane, the zwitterionic nature of the ornithine headgroup might “dilute” the negative charges on lipid A, preventing antimicrobial peptide binding to the membrane and thus limiting the permeability of cationic antimicrobial peptides (Dorrer and Teuber 1977). This resistance mechanism to antimicrobial peptides can be compared to the mechanisms described, which usually involve modifications of the membrane or surface that neutralize the negative charges in both Gram-negative and Gram-positive bacteria (Peschel 2002). These results are interesting because they suggest that the last-resort antibiotic polymyxin B could be more effective against A. baumannii and other difficult-to-treat bacteria when aminolipid biosynthesis is inhibited (Olea-Ozuna et al. 2025).

Sensing of OLs by eukaryotic hosts

The published data regarding the function of OLs in bacteria-host interaction were contradictory, and one major problem was that many different models were studied, which prevented a more general conclusion. In some (but not all) cases, mutants deficient in OL formation or modification are affected during host–microbe interactions. S. meliloti mutants deficient in OL formation form functional nodules on their host plant alfalfa (Medicago sativa) (López-Lara et al. 2005). In Brucella abortus, OLs are dispensable for the development of pathogenicity (Palacios-Chaves et al. 2011). In contrast, Agrobacterium tumefaciens mutants deficient in OL formation or OL modification show accelerated tumor formation during infection of potato tuber discs (Vences-Guzmán et al. 2013). R. tropici mutants deficient in OL hydroxylation induce the formation of nodules on their host plant, the common bean (Phaseolus vulgaris), that are defective in biological nitrogen fixation (Vences-Guzmán et al. 2011). When P. aeruginosa is cultivated in the presence of human epithelial cells, several genes are induced, including the olsBA operon (Frisk et al. 2004). It was observed that OLs increased persistence and attenuated P. aeruginosa virulence; in host cells, they reduced the production of inflammatory factors (iNOS, COX-2, PGE2, and nitric oxide) and increased intracellular Ca²⁺ release (Frisk et al. 2004). Pizzuto et al. (2024) recently showed that OLs activate TLR4 and induce NLRP3-dependent potassium efflux in primary murine macrophages and human mononuclear cells. On the other hand, in the presence of lipopolysaccharide (LPS), OLs acted as partial TLR4 antagonists, reducing LPS-induced cytokine secretion. One hypothesis is that in phosphate-depleted environments, OLs replace LPS in terms of bacterial immunogenicity, while constitutively present OLs could allow bacteria to evade the host’s immune surveillance (Pizzuto et al. 2024). This is consistent with the findings published by Kim et al. (2018), who reported that OLs increased persistence and attenuated P. aeruginosa virulence in an infection model. The observations described above for bacteria-plant interactions indicate that plants may also have a receptor that detects bacterial OL.

Presence and function of other acyloxyacyl aminolipids

Several publications, primarily from the 1970 s and 1980 s, have demonstrated that, in addition to OLs, other structural homologues with different amino acid headgroups, such as lysine, glycine, glutamine, serine-glycine, and ornithine-serine-glycine, also exist. Recently, a study by Moore et al. (2016) confirmed many of the earlier findings and detected several of these molecules for the first time using mass spectrometry (MS).

Glycine lipid (GL) was first described as cytolipin in the gliding bacterium Cytophaga johnsonae (Batrakov et al. 1999; Kawazoe et al. 1991), which is now called Flavobacterium johnsoniae. Since then, a range of mono- and di-acylated GLs and related flavolipins (FLs) with a serine-glycine headgroup have been identified in several members of the phylum Bacteroidota, including those associated with the gut and oral microbiomes, as well as oral pathogens. Two genes, glsB and glsA, encoding proteins homologous to the C- and N-terminal domains of OlsF, have been identified in different species of the genus Bacteroides. GlsB is an N-acyltransferase, and GlsA is an O-acyltransferase (Cohen et al. 2015; Lynch et al. 2017, 2019). Co-expression of the genes glsBA in E. coli leads to GL formation (Lynch et al. 2019). GlsB (BT_3459) from Bacteroides thetaiotaomicron was shown to be a glycine N-acyltransferase required for the production of both GL and FL, indicating that GL is an intermediate in FL synthesis (Lynch et al. 2019). These lipids have been referred to by different names in the literature: lyso-GL (also known as commendamide) as L342, lyso-FL as L430, GL as L567, and FL as L654. Additionally, L1256 appears to be an FL-phosphatidic acid conjugate (Guido et al. 2024).

Using GlsB from Bacteroides thetaiotaomicron (BT_3459) as a query (Lynch et al. 2019), within the phylum Bacteroidota, genes encoding GlsB homologues can be identified in the orders Cytophagales, Chitinophagales, Sphingobacteriales, Flavobacteriales, Bacteroidales, and Marinilabiliales. Genes encoding GlsB homologues have also been identified in the orders Spirochaetales, Verrucomicrobiales, and Burkholderiales in a small proportion, where the presence of GLs has not been described (Supplementary dataset S5).

In B. thetaiotaomicron, GLs have been shown to be important for resistance to bile salts and oxygen, for the transition from growth in solid to liquid medium, and for colonization of the mammalian gut, especially during early stages. A subset of bacterial lipids (including GLs, FLs, and derived molecules) produced by members of the human gut microbiota has emerged as ligands for host receptors, potentially making a significant contribution to the host–microbe dialogue. Kawai et al. (1989) demonstrated that OLs and FLs induce inflammatory immune responses, as measured by the formation of PGE2, IL-1β, and TNF-α by macrophages.

The presence of lysine lipids (LL) has been shown in A. tumefaciens, Pseudopedobacter saltans, A. baumannii, F. johnsoniae, and Rhodobacter sphaeroides (Moore et al. 2015, 2016; Olea-Ozuna et al. 2025; Tahara et al. 1976b; Vences-Guzmán et al. 2015; Zhang et al. 2009). In P. saltans, several hydroxylated versions of LL were also detected (Moore et al. 2015). LLs seem to be synthesized by genes described for OL synthesis that, in some cases, can accept lysine in addition to ornithine as an amino acid substrate. For OlsF from F. johnsoniae, and for OlsBA from A. baumannii, it has been shown that they cause the formation of LL in addition to OL (Olea-Ozuna et al. 2025; Vences-Guzmán et al. 2015).

The presence of glutamine lipids (QLs) appears to be restricted to the marine roseobacter clade, which is an important player in oceanic biogeochemical cycling (Smith et al. 2019). In Ruegeria pomeroyi, the genes glsB and olsA required for QL synthesis were identified (Smith et al. 2019). Liu et al. (2024) later suggested naming the gene encoding the first step in QL synthesis gluB (instead of glsB, as Smith et al. (2019) referred to it) to avoid confusion with glsB, which encodes the N-acyltransferase required for GL synthesis. Here, we follow this suggestion. The gene gluB encodes the OlsB homologue GluB, which catalyzes the formation of lyso-glutamine lipid (lyso-QL). Lyso-QL is then converted to QL by OlsA, which is also responsible for the formation of OL from lyso-OL in this organism. Similarly, to what has been described for OLs, QLs appear to be important for maintaining normal cellular function, replacing phospholipids under low phosphate concentrations (Moore et al. 2016; Smith et al. 2019; Zhang et al. 2009).

Bacteria belonging to the phylum Bacteroidota are the most diverse in terms of the presence of different amino acids in their headgroups. QL is the only known amino acid-containing acyloxyacyl lipid that has not been detected within the Bacteroidota phylum.

Possible applications of aminolipids

Biosurfactants are amphiphilic molecules of biological origin with surface-active properties. Their mild properties, broad pH and temperature range, and biodegradability make them suitable replacements for petrochemically derived surfactants (Desai and Banat 1997; Henkel et al. 2012). Using environmental DNA from hydrocarbon-contaminated lake soil to construct a library, the gene olsB was identified in a metagenomic screen for biosurfactant production (Williams et al. 2019). Both lyso-OL and OL showed desirable properties as biosurfactants. It is expected that other aminolipids would also exhibit interesting biosurfactant properties, but they have not been explored.

OLs have been shown to be accumulated in some bacteria in response to abiotic stresses, and in a few cases, mutants unable to synthesize OLs have been shown to be affected in abiotic stress tolerance. The genomes of E. coli strains do not encode the enzymes required for OL synthesis. Recently, in an attempt to transfer stress-resistant properties to E. coli, Bedoya et al. constructed E. coli strains expressing OlsF and OlsC, thereby forming unmodified OL and hydroxylated OL (Bedoya-Pérez et al. 2024). In a phosphate-limited growth medium, these strains reached higher cell densities than the wild type and exhibited increased tolerance to slightly acidic conditions. Given the diversity of OL-modifying enzymes, it should be possible to engineer bacterial strains with tailored membrane properties using synthetic biology.

Finally, OLs and FLs are known to be bioactive lipids. OLs were shown to interact with the TLR4/MD-2 receptor (Pizzuto et al. 2024), and it was also demonstrated that FLs are ligands for both human and mouse TLR2. The binding of FL L654 inhibits osteoblast differentiation, and L654 and the corresponding lysolipid L430 have the potential to promote TLR2-dependent bone loss in experimental periodontitis (Wang et al. 2015). L654 has also been detected in human blood samples, and patients with multiple sclerosis had significantly lower L654 levels than healthy individuals, suggesting that it could serve as a biomarker for multiple sclerosis (Farrokhi et al. 2013). Administration of L654 to mice has been shown to attenuate autoimmune disease (Anstadt et al. 2016). The FL L1256 is approximately 50-fold more potent in engaging TLR2 and the heterodimer receptor TLR2/TLR6 than the previously reported FL classes (Ryan et al. 2023). It has been suggested that OLs and FLs can serve as adjuvants and, additionally, when injected into mice before exposure to lipid A, prevent the lethal effects of bacterial endotoxin (Kato and Goto 1997; Kawai et al. 1991a, 1991b, 1999, 2000a, 2000b).

Perspectives

In recent years, our understanding of the synthesis and function of amino acid–containing acyloxyacyl lipids has advanced significantly: several new genes and pathways have been identified, and new functions and activities have been discovered, but we are still far away from a complete understanding.

Diverse aminolipid headgroups and structural modifications of these basic structures have been described, raising the question of why there is such diversity in aminolipid structures. At the same time, an open question is what makes OLs so special that three different pathways have evolved for their synthesis.

Some aminolipids (including GLs, FLs, and derived molecules) produced by members of the human gut microbiome have emerged as ligands for host receptors, potentially making a significant contribution to the host–microbe dialogue. How do they act, and what are their specific functions and activities?

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

D.M.P., M.A.V.G., and C.S. designed, drafted, and wrote the manuscript. C.S. prepared Fig. 1. D.M.P. did the computational analysis and prepared Fig. 2. All authors reviewed the manuscript.

Funding

Work in the laboratory was supported by grants from SEP-CONA(H)CyT (237713 and 425886), from Secihti (Secretaría de Ciencia, Humanidades, Tecnología e Innovación, CBF-2025-I-808), and from PAPIIT UNAM (IN202413, IN208116, IN208319, and IN211122) to CS. DMP was supported by Junta de Andalucía Predoctoral Grant (Predoc_00913).

Data availability

All datasets relevant to this manuscript or produced during the elaboration of the manuscript are included in the supplementary information submitted alongside the manuscript.

Declarations

Ethical approval

The present manuscript is a mini-review and does not involve research involving human participants or animals.

Competing interests

The authors declare no competing interests.