Introduction

The proliferation of antibiotic resistance genes (ARGs) in diverse environments has emerged as a critical global public health concern1,2. Among the various classes of antibiotics, macrolides are extensively employed in human medicine, veterinary practices, and aquaculture due to their broad-spectrum antibacterial efficacy3. Macrolides are characterized by their low biodegradability and relatively long environmental half-lives, leading to their persistent presence in water bodies4. This persistence results in continuous exposure of aquatic microbial communities to sub-inhibitory concentrations of macrolides, exerting selective pressure that promotes the development and proliferation of macrolide-resistant bacteria5,6. Furthermore, the selective pressure enhances the horizontal gene transfer (HGT) of ARGs among bacteria through mechanisms such as conjugation, transformation, and transposition, accelerating the dissemination of resistance within microbial communities7,8. Therefore, the enduring presence of macrolide antibiotics in aquatic ecosystems not only fosters the emergence of resistant bacterial strains but also facilitates the widespread dissemination of ARGs, posing significant risks to environmental and public health9.

Macrolide resistance in bacteria is predominantly mediated by three mechanisms: target site modification, active efflux, and antibiotic inactivation10. Among these mechanisms, macrolide esterases hydrolyze the macrolactone ring, rendering macrolides ineffective11. Macrolide esterases are classified into two classes based on sequence similarity: Macrolide esterase Ere-type and Est-type. The latter has been drawing increasing attention due to its prevalence in various pathogenic bacteria and its global distribution12. For instance, EstX is widely found in human pathogens and spreads through transposons, integrons, and plasmids, which facilitates its rapid dissemination across different bacterial species13. It can degrade both veterinary and clinical antibiotics, posing a significant threat to both human and animal health. Currently, research on this family of enzymes is limited, particularly regarding their transmission in the environment, their origins, and the specific factors contributing to their spread.

Cyanobacteria, the largest group of Gram-negative photosynthetic prokaryotes, serve as an important reservoir for antimicrobial resistance genes (ARGs) in aquatic environments14. Their abundance in freshwater and marine ecosystems, coupled with their ability to harbor ARGs and engage in horizontal gene transfer (HGT), makes them significant contributors to the propagation of resistance15,16,17. Cyanobacteria often form harmful algal blooms (cyanoHABs) that degrade water quality and pose public health risks18. The structural similarity of Cyanobacteria to other bacteria renders them susceptible to antibiotics’ modes of action, facilitating the acquisition and spread of resistance genes19. For example, the sul1 gene, which confers resistance to sulfonamides, has been detected in Cyanobacteria and subsequently transferred to pathogenic bacteria in aquatic environments20. Similarly, the tetA gene, responsible for tetracycline resistance, has been identified in Cyanobacterial strains and shown to contribute to the spread of resistance among other bacterial species21. Cyanobacterial blooms exacerbate the dissemination of ARGs, creating a feedback loop that further promotes resistance gene spread22. Given their widespread presence and ecological significance23, Cyanobacteria represent a serious concern as reservoirs of resistance24, necessitating further research into their role in ARG dissemination both in laboratory and natural settings. Whether macrolide esterases, such as EstX, exist and are propagated in Cyanobacteria remains an open question that warrants further investigation.

This study aims to address these gaps by investigating the role of Cyanobacteria in the dissemination of macrolide resistance genes. Specifically, we seek to: (1) screen and identify potential macrolide resistance genes in Cyanobacteria; (2) biochemically characterize identified macrolide esterases and elucidate their mechanisms of action; and (3) assess the potential for horizontal transfer of macrolide resistance genes within Cyanobacterial populations. The findings of this research will provide critical insights into the ecological dynamics of macrolide resistance, highlighting the significance of Cyanobacteria as reservoirs and conduits for the spread of resistance. This knowledge is essential for developing informed strategies to mitigate the public health impact of antibiotic resistance in aquatic environments.

Results

Screening of macrolide esterase genes

Cyanobacteria were known for its high susceptibility to macrolide antibiotics, even at low concentrations (The EC50 of Cyanobacteria for macrolide antibiotics is only 0.0121 mg/L19). Despite this susceptibility, there are no reports studying macrolide resistance genes in biochemical level or molecular level in Cyanobacteria. To identify potential macrolide resistance genes in cyanobacteria, we performed a systematic search of the NCBI genome database, encompassing 100 cyanobacterial species (totaling 18,839 genomes). Using the sequences of known esterases EstT and EstX as queries, we successfully identified three potential macrolide esterases (Table S1). They were designated as NOD-1 from Nodosilinea sp. LEGE 07298, OCA-1 from Oscillatoriales cyanobacterium C42_A2020_001 and OCB-1 from Oculatellaceae cyanobacterium Prado106. A sequence comparison of these three proteins with the known esterases EstT and EstX was performed. The results revealed sequence identities ranging from 44.13% to 47.14% with EstT, and from 60.36% to 62.37% with EstX (Fig.1B). Surprisingly, no candidate proteins were found to share similarity with Ere-type macrolide esterases.

Fig. 1: Phylogenetic analysis, sequence similarity, and functional characterization of three macrolide esterases from cyanobacteria.
figure 1

A Phylogenetic tree of esterases NOD_1, OCA_1 and OCB_1 with other similar enzymes. And plotted with the Chiplot online website (https://www.chiplot.online/). B This heat map depicts the protein sequence similarities between the three mined proteins. The sequence similarities between the novel esterases and known macrolide esterases range from 44.13% to 62.37%. C Antimicrobial susceptibility tests of three recombinant bacterias, E.coli BL21-pET28a-NOD_1, E.coli BL21-pET28a-OCA_1, and E.coli BL21-pET28a-OCB_1 showed their susceptibility to 12 macrolide antibiotics. Three recombinant strains exhibited resistance to tylosin. The experiment was performed in triplicate.

Full size image

To further analyze the evolutionary relationship between predicted macrolide esterases from Cyanobacteria, phylogenetic tree analysis was conducted. The phylogenetic analysis shows that three of these identified proteins cluster into one clade that close to EstX and EstT. The macrolide esterases that have been identified are mainly derived from Bacteriodetes and Pseudomonadota, while the macrolide esterases from the three Cyanobacteria have been identified and are more closely related to the Pseudomonadota (Fig.1A). Given that the sequence identity between these three proteins and the known protein EstT is lower than 60%, we hypothesize that they may share similar functions.

The three characterized esterases exhibit broad taxonomic diversity, originating from phylogenetically distinct Cyanobacterial strains: Nodosilinea sp. (NOD-1), an unclassified Oscillatoriales cyanobacterium (OCA-1), and a member of the Oculatellaceae family (OCB-1). This taxonomic distribution spans multiple Cyanobacterial orders, underscoring the widespread evolutionary acquisition or conservation of macrolide-inactivating capacity within this phylum. Additionally, the relatively long branch lengths of these proteins in the phylogenetic tree suggest they have undergone evolutionary divergence (Fig.1A). The sequence similarity heatmap further supports this observation, showing low sequence identity among the three proteins. For example, the highest sequence identity is between OCA-1 and OCB-1, up to 74.56%, while the lowest is between OCB-1 and EstT, and the similarity is 44.13% (Fig.1B). This indicates that the three proteins are diverse and representative.

Phylogenetic analysis revealed that the cyanobacterial esterases identified in this study (NOD-1, OCA-1, and OCB-1) form a distinct, well-supported clade with the known esterases EstX and EstT. This suggests they may share a relatively recent common ancestor and are likely functionally similar, possessing the potential to degrade 16-membered-ring macrolide antibiotics. Based on this evolutionary evidence, we hypothesize that these three uncharacterized cyanobacterial proteins are potential macrolide esterases, and we subsequently conducted functional characterization studies to test this hypothesis.

A combination of phylogenetic proximity to established macrolide esterases (EstX and EstT) (Fig. 1A), significant overall amino acid sequence identity with these reference esterases, and contain the conserved serine hydrolase catalytic triad (Ser-His-Asp/Glu) characteristic of macrolide esterases. Figure S1 led us to propose that these three uncharacterized cyanobacterial proteins are putative macrolide esterases. We selected them for functional characterization to test this hypothesis.

Phenotype of Escherichia coli containing macrolide esterase

Strains carrying the three resistant genes showed the tendency of increased resistance. Moreover, three of the recombinant strains showed increased MIC values for tylosin among all 12 tested macrolide antibiotics. Recombinants carrying OCA-1 and OCB-1 exhibited approximately a 4-fold increase in tylosin MIC compared to the BL21 parent strain, while NOD-1 showed a 2-fold increase. These results indicate that these three proteins can degrade the antibiotic (Fig.1C and Table S2).

Next, we investigated the degradation abilities of the three esterases against macrolide antibiotics using the disk diffusion assay. The three esterases that showed phenotypes in the MIC assay successfully degraded tylosin, resulting in the disappearance of inhibition zones. Notably, OCA-1 demonstrated the broadest substrate spectrum, degrading five antibiotics: tylosin, tilmicosin, spiramycin, tildipirosin, and leucomycin A1 (Fig.2B). Intriguingly, substrate preferences correlated with antibiotic applications: NOD-1 exclusively degraded veterinary-use antibiotics (tylosin, tildipirosin, and tilmicosin) (Fig.2A), whereas OCA-1 and OCB-1 additionally exhibited degradation activity against human-use antibiotics (spiramycin and leucomycin A1) (Fig.2B–D). Based on the number of antibiotics degraded, the order of effectiveness was OCA-1 > OCB-1 > NOD-1 (Fig.2D). OCA-1 had the broadest substrate range, degrading five macrolides, including spiramycin and leucomycin A1 (human-use), as well as tylosin, tilmicosin, and tildipirosin (veterinary-use) (Fig.2B, D).

Fig. 2: Zone of bacterial inhibition produced by the reaction of three newly excavated enzymes with 12 macrolide antibiotics.
figure 2

A–C The effects of three new enzymes NOD_1, OCA_1, and OCB_1 on the growth of Staphylococcus aureus after reaction with antibiotics, and the change of the size of the inhibition zone can reflect their ability to degrade antibiotics. The experimental strain was ATCC25923. D Heat map of NOD_1, OCA_1, and OCB_1 resistant phenotypes, with gray indicating non-resistance and pink indicating resistance. All data are given as mean ± SD, n = 3.

Full size image

Expression and purification of esterases

To further investigate the biochemical properties of the esterases, we attempted expression and purification. Among the three proteins, only OCA-1 was expressed and successfully purified, while the other two were expressed as inclusion bodies, or there was no way to obtain sufficient amounts of enzymes for subsequent analysis (Fig. S2). SDS-PAGE analysis confirmed successful purification, showing a single band at approximately 31 kDa, consistent with the predicted molecular weight of OCA-1 (31.34 kDa) (Lane 2, Fig. 3A). The purity of OCA-1 exceeded 90%, making it suitable for further characterization studies.

Fig. 3: Purification and Enzymatic Properties of Esterase OCA_1.
figure 3

A SDS-PAGE of purified esterase OCA_1. Lane 1 is the marker, and lane 2 is the purified esterase. The gel was stained with Coomassie brilliant blue. The effects of temperature and pH on the OCA_1 activity of recombinant enzymes were investigated, and the substrates were p-nitrophenyl butyrate (B, C). B The corresponding enzyme activity of the OCA_1 at 0–100 °C was determined separately, and the maximum enzyme activity measured was defined as 100%, which was used to calculate the relative enzyme activity. Data are given as mean ± SD, n = 3. C Effect of pH on OCA_1 enzyme activity. The relative activity is calculated by measuring the change in enzyme activity in the pH range between pH 3 and 12 at 40 °C, and the maximum enzyme activity measured is defined as 100%. Data are given as mean ± SD, n = 3.

Full size image

The results from the disk diffusion assay were not fully consistent with those from the broth microdilution method (Figs. 1C and 2). For example, NOD-1 could degrade Tylosin, OCA-1 could degrade 5 antibiotics (tylosin, tilmicosin, spiramycin, tildipirosin, and leucomycin A1), and OCB-1 could degrade 4 antibiotics (tylosin, tilmicosin, spiramycin, and leucomycin A1), but only tylosin showed changes in MIC. We hypothesize that this discrepancy may stem from factors such as gene copy number and promoter strength in the heterologous host, which likely differ from expression levels in the native bacterial context.

We also measured the inhibition zone size over time after OCA-1 reacted with five antibiotics (Fig. 4). After 30 min of reaction with tylosin, no inhibition zone was observed, indicating that OCA-1 completely degraded tylosin within 30 min. As shown in the figure, it took 2 h to fully degrade tilmicosin and 4 h to degrade tildipirosin. However, even after overnight reactions, spiramycin and leucomycin A1 were not completely degraded. The degradation rate follows the order: tylosin > tilmicosin > tildipirosin > spiramycin > leucomycin A1.

Fig. 4: Analysis of the time-dependent degradation of five macrolide antibiotics by esterase OCA_1.
figure 4

A The temporal degradation curves of the five macrolide antibiotics were OCA_1, reflecting the difference in the strength and weakness of esterases in the degradation of different types of antibiotics. The highest degradation rate was observed for tylosin, followed by tilmicosin. Data are given as mean ± SD, n = 3. B. Analysis of the plate inhibition zones of five macrolide antibiotics against Staphylococcus aureus ATCC 25923 following hydrolysis by OCA-1.

Full size image

Enzymatic properties

OCA-1 exhibited high enzymatic activity25,26. between 20 °C and 50 °C, but was completely inactivated at 60 °C, with an optimal temperature of 40 °C (Fig. 3B). In pH tests, OCA-1 retained high activity in the pH range of 6–7.5. However, as the pH increased beyond this range, its activity decreased until complete inactivation, indicating that OCA-1 is sensitive to extreme pH levels and performs best in a neutral environment (Fig.3C). The kinetic parameters of OCA-1 for the substrate p-NPB were determined, with a Michaelis constant (Km) of 0.29 mg/mL and a maximum velocity (Vmax) of 5.02 μmol/min·mL (Figure S3).

The hydrolysis was further confirmed by tandem mass spectrometry. Notably, all five antibiotics exhibited a consistent +18.0 Da mass shift (Δm/z) following OCA-1 treatment, precisely matching the theoretical mass increment from water addition (H₂O: 18.0 Da). For example, tylosin’s molecular weight changed from 916.4 m/z to 934.4 m/z, tilmicosin changed from 869.3 m/z to 887, tildipirosin changed from 734.4 m/z to 751.9, spiramycin changed from 842.8 m/z to 861.3, leucomycin A1 changed from 786 m/z to 804 m/z. This increase of 18 Da likely corresponds to the addition of a water molecule, suggesting that OCA-1 may act as a macrolide esterase by hydrolyzing ester bonds (Fig.5A-E).

Fig. 5: Mass spectra of five antibiotics before and after degradation by protein OCA_1.
figure 5

A Mass spectra of spiramycin (m/z 843.05) before and after degradation. B Mass spectra of tilmicosin (m/z 869.15) before and after degradation. C Mass spectra of tildipirosin (m/z 734.02) before and after degradation. D Mass spectra of tylosin (m/z 916.10) before and after degradation. E Mass spectra of leucomycin A1 (m/z 785.96) before and after degradation.

Full size image

Catalytic mechanism analysis

We used ColabFold to predict the structures of three proteins. The predicted pLDDT scores ranged from 95.5 to 97.3 (Table S3), indicating that the predictions are relatively reliable (Fig. 6A, B). The overall structures and binding pockets of the three proteins are highly similar (Fig. 6B, C and Table S4). Given that the catalytic mechanism of the macrolide esterase from Cyanobacteria may differ, we aimed to further identify its catalytic amino acids. By comparing the amino acids in the binding pocket, we found that the conservation level is high, with 28 out of 284 amino acids being fully conserved. The binding pocket contains 2 serines, of which only S34 and S102 are conserved (Fig. 6D). Considering that such amino acids are often spatially close to the ester bond of macrolides, we performed molecular docking for further screening. The results showed that S102 is the closest to the ester bond, suggesting that S102 is the key catalytic residue in OCA-1 (Figs. S4–5 and Table S5).

Fig. 6: Comprehensive structural analysis of three novel esterases and functional validation of site-directed mutant.
figure 6

A The 3D structure of Phylogenetic tree of three newly mined proteins, plotted using the chiplot online website (https://www.chiplot.online/). B proteins NOD_1, OCA_1, and OCB_1 expected three-dimensional structures. The colors in the model represent the predicted local distance difference test (pLDDT) confidence scores, with very high (>90), high (80), medium (70), low (60), and very low (<50) levels depicted in red, yellow, green, light blue, and blue, respectively. C The binding pockets of protein NOD_1, OCA_1, and OCB_1 are displayed, and the yellow area is the binding pocket site predicted by PrankWeb. D Sequence identification maps of recombinant proteins NOD_1, OCA_1, and OCB_1. The height of the letters indicates the base frequency at each location, and the abscissa represents the site of the amino acid. This figure was generated using WebLogo 3 (https://weblogo.threeplusone.com/). E SDS-PAGE of recombinant bacteria E. coli BL21-pET28a- OCA_1 mutant strain. Lane 1 was marker, and lane 2 was the protein of the E. coli BL21-pET28a- OCA_1 mutant strain purified by nickel column, and the band size was 31 kd, which was consistent with the expected protein size. F The histogram showed whether the mutant strain protein of OCA_1 had a degrading effect on 12 macrolide antibiotics. Data are given as mean ± SD, n = 3. G It reflects the susceptibility of the recombinant bacterium E. coli BL21-pET28a-OCA_1-S102 mutant to 12 macrolide antibiotics. The experiment was performed in triplicate.

Full size image

Using the purified S102A mutant (Fig. 6E), we performed antibacterial zone of inhibition assays and found that the mutant lost its ability to hydrolyze other macrolide antibiotics (Fig. 6F). Additionally, the MIC of E. coli BL21-pET28a carrying the S102A mutant was identical to that of the empty vector control (Fig. 6G). This confirms that S102 is the key catalytic amino acid, consistent with previous reports.

Analysis of genomic context

The confirmation of macrolide resistance genes in Cyanobacteria prompted us to investigate whether horizontal transfer of these macrolide esterase genes could occur. Genomic context analysis13 revealed that these esterase genes are conserved and distributed across 14 Cyanobacterial species derived from diverse habitats—including hot springs, wetlands, and terrestrial environments (Fig. 7B). Notably, the genomic regions surrounding the macrolide esterase genes were found to be enriched with mobile genetic elements and other antibiotic resistance determinants, such as IS family transposases27, Rpn family recombination-promoting nuclease28, and aminoglycoside nucleotidyltransferase29. Furthermore, comparative analysis indicated that the genetic backgrounds of macrolide esterases from geographically distant sources—such as a Chinese strain (GCF_014695395.1) and a Slovakian strain (GCA_019358835.1)—display remarkable similarity. Strikingly, conserved genetic contexts were also identified among different Cyanobacterial genera within the same geographic region. For example, macrolide esterases from Chinese Oculatella sp. (GCF_014696015.1) and an unclassified Cyanobacteria bacterium (GCA_014695345.1) share a similar genomic neighborhood, as do those from Chinese Phormidium tenue (GCF_014696675.1) and Nodosilinea sp. (GCF_014695395.1). (Fig. 7A)

Fig. 7: Gene cluster propagation map.
figure 7

A Genomic context linkage analysis of macrolide resistance gene loci in representative Cyanobacteria species. The genomes of Cyanobacteria derived from OCA-1 with more than 80% protein similarity were selected for comparative analysis, and the upstream and downstream regions of macrolide resistance genes were compared and mapped. Purple arrows represent the macrolide resistance gene from Cyanobacteria species. If two sequences from two loci share sequence similarity, they are linked with shadows that change color according to sequence similarity. B Global distribution of macrolide resistance genes in cyanobacteria. Countries where these genes have Appeared are highlighted in red, and those without are shown in gray. The highest prevalence was observed in Portugal and China.

Full size image

Discussion

In this study, we identified three novel functional macrolide esterases—NOD-1, OCA-1, and OCB-1—from cyanobacteria by a systematic screening of the NCBI genome database. Biochemical characterization confirmed that these enzymes operate through a key serine residue (S102) and specifically hydrolyze 16-membered ring macrolide antibiotics. Moreover, genomic context analysis indicated association of these esterase genes with various mobile genetic elements, pointing to a potential risk of horizontal gene transfer and dissemination.

Phylogenetic analysis revealed that the three esterases form a distinct clade, distantly related to the Bacteroidota-derived EstT but more closely aligned with EstX from environmental Pseudomonadota. This finding underscored the complexity and underappreciated diversity in the evolution and distribution of macrolide resistance genes in the environment30. BLAST analysis against the NCBI database did not identify homologous sequences of these novel genes in the Bacteroidetes and Pseudomonadota phyla (Tables S6 and S7). Notably, this work provided the first systematic identification and functional validation of such esterases in cyanobacteria—the globally pivotal primary producers. Cyanobacteria are widely distributed in both freshwater and marine ecosystems31. They exhibit a high sensitivity to macrolide antibiotics19. For instance, experiments have shown that Cyanobacteria are antibiotic-sensitive organisms, and their EC 50 value against macrolide antibiotics is less than 0.03 mg/L19. Given the widespread contamination of macrolide antibiotics in the environment32,33, Cyanobacteria are likely exposed to sub-inhibitory concentrations over long periods, experiencing sustained selective pressure. This selective pressure may drive the evolution of adaptive mechanisms in response to these antibiotics34. Our study not only revealed a potential mechanism by which Cyanobacteria may degrade antibiotics through macrolide esterases but also provided new insights into their role as reservoirs of antibiotic resistance genes in the environment.

Biochemical assays indicated that the esterases generally exhibit pronounced hydrolytic activity against 16-membered ring macrolides (e.g., tylosin and tilmicosin). By contrast, no activity was observed against 14- or 15-membered macrolides, such as erythromycin, roxithromycin, or azithromycin. This substrate specificity mirrored, yet remains distinct from, those of known esterases EstT and EstX13,35. We proposed that variations in catalytic residues within the substrate-binding pocket underlied the divergent substrate profiles. Additionally, OCA-1 showed optimal activity at pH 8.0 and 40 °C, consistent with its isolation from a neutral-alkaline aquatic habitat, suggesting environmental adaptation of its enzymatic function36.

The complete inactivation of the S102A mutant aligns with the canonical catalytic mechanism of the α/β-hydrolase superfamily37,38. Prior work on EstX and EstT has similarly established the essential role of the catalytic serine13,35, and our study corroborates this conserved mechanism in cyanobacterial homologs. A key distinction of our work, however, lies in the low sequence similarity and divergent substrate profiles of the cyanobacterial esterases (e.g., OCA-1 and OCB-1) relative to EstT and EstX. For instance, EstX degrades tylosin, tidipirosin, tilmicosin, and leucomycin A513, EstT targets tylosin, tilmicosin, and tildipirosin substrates35, and the Cyanobacterial enzymes identified in this study can degrade tylosin, tilmicosin, spiramycin, tildipirosin, and leucomycin A1 substrates. Given that all these enzymes utilize the same catalytic triad, it is unlikely that the triad itself accounts for the differences in substrate specificity. When we examined the overall three-dimensional structure of these proteins, we found that Cyanobacterial esterases share a high degree of spatial similarity in their three-dimensional structures with EstX and EstT, as indicated by DALI scores ranging between 45.9 and 54.2 (Table S8). Therefore, overall protein structure also does not appear to explain the observed differences in substrate specificity. Considering that, many studies suggest that amino acids within the binding pocket may significantly influence substrate preferences. In our analysis, we found notable differences in the conserved amino acids surrounding the binding pocket between Cyanobacterial esterases and EstX/EstT (Table S4). For example, S102 amino acid determine the entry of the binding pocket. These differences suggest that, despite a conserved catalytic core, these cyanobacterial esterases may represent a novel functional subfamily with an evolutionary path distinct from EstX. Confirming this might need further experimental evidence, and important evidence for understanding the change of drug resistance genes from narrow spectrum to broad spectrum.

Antimicrobial resistance (AMR) poses severe global health and economic threats39. Rising consumption of macrolide antibiotics40,41,42, has increased their environmental burden32,33, sustaining selective pressure that promotes the maintenance and spread of resistant phenotypes43. Addressing macrolide resistance requires identifying environmental reservoirs and transmission mechanisms of relevant resistance genes. Accordingly, we focused on cyanobacteria—a highly abundant and widely distributed phylum—to mine macrolide resistance genes and elucidate their functional mechanisms, distribution, and transfer potential.

Genomic context analysis offers critical insights into resistance gene dissemination risk44. Within the cyanobacterial order Nostocales, we observed the highest frequency of predicted AMR genes (26%; Fig. 7A). Timms et al. previously identified predicted resistance genes in 23% of Nostocales genomes, describing this group as highly divergent, bloom-forming, and rich in AMR genes, often coexisting with human and animal pathogens14. Earlier evidence also suggests that Nostoc linckia can develop resistance to antibiotics such as penicillin via nuclear or gene mutations45. Furthermore, Timms et al. proposed that cyanobacterial resistance to ansamycins may represent an ancient adaptation facilitating symbiosis with actinobacteria45. As actinobacteria are Gram-positive and naturally susceptible to macrolides, cyanobacteria in such symbiotic associations may experience indirect antibiotic pressure, potentially favoring the acquisition and retention of resistance genes.

In several instances (Fig. 7A), we detected flanking mobile genetic elements and other resistance genes near the target esterase genes. For example, the genome of Desmonostoc muscorum (GCF_015207005.2) was flanked by insertion sequence (IS) elements, strongly implying the potential for horizontal transfer of these macrolide esterase genes within microbial communities46,47. It should be noted, however, that this study highlights genetic potential, not confirmed dissemination events.

According to a compilation by Li et al. on global macrolide antibiotic concentrations in natural waters (2000–2022)48, regions with higher contamination levels—such as China and Portugal—also host cyanobacteria carrying macrolide resistance genes (Fig. 7B). This correlation further supports the link between antibiotic residue exposure and enrichment of resistance genes49.

Although we have only found about 10 macrolide esterase genes derived from Cyanobacteria, this is mainly due to the fact that the number of genomes of Cyanobacteria in the current database is much lower than that of bacteria, and the diversity is relatively small. Given the widespread presence and significant biomass of Cyanobacteria in various natural environments50, the number of Cyanobacteria carrying macrolides may be very high. The risk is particularly alarming in the context of increasing Cyanobacterial blooms. These blooms, which are becoming more frequent due to climate change and eutrophication51,52,53,54, create ideal conditions for the proliferation and dissemination of resistance genes55. The spread of such genes to pathogenic bacteria in aquatic environments could further exacerbate the global challenge of antibiotic resistance. This highlights the urgent need for enhanced monitoring and deeper investigation into the ecological and evolutionary dynamics of these resistance mechanisms, as well as their broader implications for environmental and human health.

Methods

Screening of macrolide esterases from Cyanobacteria

To investigate the presence of macrolide esterases in cyanobacteria, we conducted a systematic bioinformatic analysis based on the NCBI GenBank database. First, genomic data from 100 cyanobacterial species (covering 18,839 genomes in total) were collected. Subsequently, a local homology search was performed using Diamond with the known macrolide esterases EstX and EstT as query sequences, with the parameters set as follows: E-value threshold of 1e-10, a minimum sequence identity of 40%, and a sequence coverage of at least 80%. From the obtained homologous sequences, we selected the three most frequent representative sequences for further study. To resolve their evolutionary relationships, a phylogenetic tree was constructed using these candidate esterase sequences, the experimentally verified EstT and EstX, and outgroup proteins (Aclacinomycin methylesterase RdmC, Q54528.1, and Rhodomycin D methylesterase DnrP, Q54809.1/Q55217.1). Finally, the phylogenetic tree was visualized using ChiPlot.

Gene synthesis and protein expression, and purification

The gene encoding the target protein with an N-terminal 6X-His tag was synthesized by Jiangsu Genecefe Biotechnology and cloned into the pET28a expression vector. The resulting recombinant plasmid was transformed into Escherichia coli BL21(DE3), and colony PCR and Sanger sequencing were utilized to verify the correctness of the construct (Table S9). The positive clones were selected using kanamycin (100 µg/mL).

The recombinant E. coli strain carrying the target gene was cultured at 37 °C in LB medium supplemented with 100 µg/mL kanamycin. When the culture reached the logarithmic growth phase (OD600 between 0.6 and 0.8), protein expression was induced by adding isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.5 mM. The culture was then incubated at 25 °C for 16-18 hours to allow protein expression. After induction, cells were harvested by centrifugation and resuspended in Buffer A (20 mM Tris, 500 mM NaCl, pH 7.0). The cells were lysed by sonication, and the lysate was centrifuged to collect the supernatant.

The supernatant containing the soluble protein was loaded onto a nickel affinity chromatography column for purification. The column was washed with varying concentrations of imidazole to remove non-specifically bound proteins. The purified fractions were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) to determine protein molecular weight. The eluted protein was then pooled and dialyzed against 20 mM Tris-HCl buffer (pH 7.0) for further analysis.

Antibiotic susceptibility test

The susceptibility of Escherichia coli BL21(DE3) strains carrying different plasmid constructs against 12 macrolide antibiotics, including 14-membred ring (erythromycin, roxithromycin, and clarithromycin), 15-membred ring (azithromycin), 16-membred ring (josamycin, tilmicosin, spiramycin, leucomycin A1, midecamycin, tildipirosin, acetylspiramycin, and tylosin) was determined using the broth microdilution method. Antibiotics were serially diluted in Mueller-Hinton (MH) broth using a twofold dilution series, and the concentration of antibiotics were 64 (erythromycin), 128 (roxithromycin), 32 (clarithromycin), 2 (azithromycin), 256 (josamycin), 512 (tilmicosin), 256 (spiramycin), 512 (leucomycin A1), 512 (midecamycin), 2 (tildipirosin), 512 (acetylspiramycin) and 1024 μg/mL (tylosin) respectively. A standardized bacterial inoculum was then added to each well containing the diluted antibiotics. The cultures were incubated at 37°C for 24 hours, after which the optical density at 600 nm (OD600nm) was measured. The lowest concentration of the antibiotic that inhibited visible bacterial growth was recorded as the minimum inhibitory concentration (MIC) for each tested strain.

ESI-MS analysis

To evaluate the enzymatic activity of OCA-1 on macrolide antibiotics, 200 µL of crude enzyme extract was mixed with 800 µL of 10 µg/mL macrolide antibiotic solution (for tildipirosin, 30 µg/mL) and incubated overnight at 37 °C. The reaction was terminated by boiling for 5 min. A control reaction was performed by incubating heat-inactivated enzyme with the antibiotic. The resulting reaction mixtures were divided into two portions: one for electrospray ionization mass spectrometry (ESI-MS) analysis to investigate whether OCA-1 could degrade the macrolide antibiotics, and the other for an agar diffusion (inhibition zone) assay to evaluate the effect of the reaction mixture on the growth of Staphylococcus aureus ATCC 25923.

To assess the degradation of the macrolide antibiotic, the reaction mixture was analyzed by electrospray ionization mass spectrometry (ESI-MS). The analysis was performed under the following optimized parameters: a scan range of 100–1000 m/z, curtain gas pressure of 40 psi, ion source gases 1 and 2 both set to 30 psi, and a drying gas temperature and flow rate of 450 °C and 5 L min⁻¹, respectively. The ESI voltage was configured at 5500 V in positive ion mode (ESI+) and 4500 V in negative ion mode (ESI-).

Inhibition zone analysis

Inhibition zone analysis was performed using the agar well diffusion method as previously described13. This assay served as a preliminary evaluation of the antibacterial activity of the reaction mixtures. It is suitable for initial screening of antimicrobial properties, and the assay was displayed by the standard of CLSI 2024.

Briefly, Luria–Bertani Agar (LBA) was cooled to 35–45 °C and inoculated with a standardized suspension of Staphylococcus aureus. The inoculated medium was mixed thoroughly, and 10 mL was dispensed into each sterile 90 mm Petri dish. After solidification, 4.5 mm diameter wells were aseptically punched into the agar using a sterile cork borer. Then, 30 µL of each reaction mixture was added into the respective wells. The plates were incubated at 37 °C for 18–24 h, after which the diameters of the inhibition zones were measured in millimeters to assess antibacterial activity.

Enzyme activity assay

Enzyme activity is defined as the amount of enzyme required to produce 1 μmol of degradation product per minute under optimal conditions. For assays using p-nitrophenyl butyrate (p-NPB) as the substrate, the amount of p-nitrophenol released is measured. To perform the assay, 100 µL of purified enzyme solution was added to 900 µL of substrate mixture (containing 2 mM p-NPB, 0.1% Triton X-100, and 20 mM Tris-HCl, pH 7.0). The reaction was carried out at 37 °C for 20 min, after which the reaction was stopped by boiling for 5 min. The absorbance of the reaction mixture was measured at 405 nm using a UV spectrophotometer to determine the enzyme activity.

Determination of optimal temperature, pH, and kinetic parameters

The optimal temperature for the purified macrolide esterase activity was investigated using p-nitrophenyl butyrate (p-NPB) as the substrate. A reaction mixture containing 100 µL of purified purified macrolide esterase enzyme and 900 µL of substrate solution (2 mM p-NPB, 0.1% Triton X-100, and 20 mM Tris-HCl, pH 7.0) was incubated at various temperatures (0 °C, 10 °C, 20 °C, 30 °C, 40 °C, 50 °C, 60 °C, 70 °C, 80 °C, 90 °C, and 100 °C) for 20 min. The reaction was terminated by boiling for 5 min. A heat-inactivated enzyme was used as the control. (OCA-1 lost its activity upon incubation in boiling water for 1 h.)

To determine the optimal pH, the same protocol was followed, keeping the temperature constant. To evaluate the effect of pH on enzyme activity, the enzyme and substrate mixture were incubated in buffers of varying pH (3–12), which were prepared using the tribasic acid and 0.2 M NaOH. All experiments were performed in triplicate, unless otherwise specified. The kinetic parameters of the macrolide esterase OCA-1 were determined under optimal conditions (pH 7.0, 40 °C). The enzyme was mixed with varying concentrations of the substrate, and the initial reaction rates were measured.

Structural prediction and molecular docking

In this study, the ColabFold platform was used to predict the 3D structure of the target protein, and the predicted PDB files were obtained with parameters number_recycles 50, max_msa 512-1024, number_relax 5, rank_plddt, and number_models 556. PyMOL57 was then used for visual analysis, and prankweb was used to predict the binding pockets of the NOD-1, OCA-1, and OCB-1 models. OCA-1 was docked with macrolide antibiotics (tylosin, tildipirosin, tilmicosin, spiramycin, and leucomycin A1) using Discovery Studio 2019, and the optimal binding site was predicted based on the size and distance of binding energy.

Distribution and genetic background of macrolide esterases in cyanobacteria

We investigated the distribution of cyanobacterial esterases across bacterial genomes using the NCBI bacterial genome database. Briefly, a total of 1.3 million bacterial genomes were downloaded from the NCBI Reference Sequence Database (July 2023), and the protein sequences from each genome were compiled into a FASTA file. Subsequently, protein sequence similarity analysis was performed using Diamond with OCA-1 as the query sequence. Through this process, we identified bacterial genomes that had OCA-1 (100% identity and 100% coverage). For each identified genome, we also compiled associated metadata, including host and geographical location.

To investigate the genomic context of cyanobacterial macrolide esterases, we systematically extracted and analyzed the ten protein-coding sequences flanking each esterase gene in the upstream and downstream regions. Functional annotation of these sequences was performed using the NCBI and CARD databases. The linkages between various representations of these analyses were visualized using the clinker online website (https://cagecat.bioinformatics.nl/).

Statistical analysis

The creation of graphs and the analysis of significance were performed using Adobe Illustrator 2021 and GraphPad Prism (version 9.5.1). Significance was denoted by a P value below 0.001, signifying differences between the treated and control groups. Data were presented as means ± standard deviation (SD), and all experiments had a minimum of three biologically independent replicates.