Phenotypic resistance profiles and resistome variations between endophytic and epiphytic bacteria in apple fruits

Abstract

Background

In recent years, there has been increasing concerns about antibiotic resistance. Although studies have investigated resistance in food-associated bacteria, fresh produce microbes remain underexplored as potential hub of resistance genes capable for horizontal transfer to human via consumption. To this end, we tested the antibiotic resistance profiles of bacterial isolates recovered from Golden Reinders and Mandy apple cultivars. We aimed to investigate the effects of orchard-cultivar combinations and microbial lifestyle on the antibiotic resistance profiles. The apples (Golden Reinders and Mandy) were sampled from four separate orchards (EEL-Lleida, Esterri, Fruits-de-Ponent and Gotarta) in Spain. We used combination of culture-dependent and whole genome sequencing approaches to analyse the antibiotic resistance profiles.

Results

A total of 516 bacterial isolates were screened for susceptibility against seven different classes of antibiotics. Results showed that 272 isolates were resistant to at least one antibiotic. From those, 203 were epiphytes and 95 classified as endophytes (isolated from surface-sterilized apple peels), whereas 26 isolates were shared between the groups. The resistance profiles varied across the antibiotics, with over 50% of the isolates exhibiting resistance to tetracycline, quinolones and cephalosporins. In contrast, none of the isolates showed resistance to imipenem. Whole genome sequencing (WGS) was performed on 18 isolates, however, only 10 genomes passed quality-control thresholds and were included in subsequent resistome analyses. We found ARGs encoding resistance to 14 main antibiotic classes, with the majority of the confirmed resistances attributed to multidrug resistance (MDR). Only few target-specific ARGs were annotated, including (Rif)iri (rifampicin), lnu(A) (lincomycin) and FosD (Fosfomycin). Pantoea agglomerans possessed higher number of ARGs, while Staphylococcus arlettae exhibited notable prevalence of plasmid-encoded ARGs.

Conclusion

Overall, the study highlights the prevalence of antibiotic resistance in apple microbiomes. The presence of multidrug-resistance (MDR) genes further underscores the persistent threat of ‘antibiotic resistance’, underlining the necessity for deeper insight into antibiotic resistance within food chain.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40793-026-00880-0.

Background

Antibiotics are specifically a type of antimicrobial agents that either destroy or inhibit bacterial growth [1]. Initially, most of the antibiotics were developed to treat human diseases; however, their usage was rapidly extended to treat and prevent animal infection, and in some cases plant diseases [2, 3]. However, resistance to almost all the existing antimicrobial agents threatens to erode the gains so far in combating infectious diseases. Loss of antibiotics effectiveness and increased cases of incurable illnesses in hospitals have driven antimicrobial reistance (AMR) to the top global health agenda [4]. Despite the fact that majority of the resistant strains are documented in clinical settings and often associated with certain disease infections, human colonization with antibiotic resistant bacteria (ARB) can occur independently outside healthcare environments [5–7]. In fact, plant-associated microbial communities intrinsically harbor diverse resistomes including genes with potential clinical relevance. This was shown for the phyllosphere of native plants [8] as well as leafy greens [9–11]. However, food resistomes, in comparison to other environments associated with humans, are rarely studied [12] despite their impact on health issues [13].

Food microbiomes are vital when considering food safety [14], shelf life [15] and fermentation processes [16], and it has been highlighted as the “missing link” [17]. Even though health-beneficial microorganisms associated with food have been identified [18], studies showed that fruits and vegetables [19] represent an important human–environment interface and can serve as an entry point for AMR to humans [9–11, 20]. The primary concern is the ingested pathogenic bacteria that harbor antibiotic resistance genes (ARGs). The ability of bacteria to transfer ARGs to other bacterial species through horizontal gene transfer (HGT) suggests that even commensal species may be significant vectors potentially increasing the risk of ARGs spreading to human microbiomes [21–23]. Fresh produce harbors diverse and dynamic microbial communities, which include not only plant microflora but also plant saprotrophs, antagonists and pathogens [24, 25]. These microbial communities have co-evolved with plants [26, 27], some of which inherently possess resistance mechanisms as a survival strategy. Fresh produce resistomes varies depending on several factors, including plant species, irrigation practices, wild or domestic animals, postharvest management and environmental conditions [11, 28]. The resistance traits may arise from environmental selective pressures, competition for resources, exposure to plant secondary metabolites or adaptation to abiotic conditions [29–31]. Moreover, soil microbiomes serves as vast reservoirs for AMR [32]. Since plants with its rich portfolio in (antimicrobial) secondary metabolites interact with soil microbiota, ARGs may be naturally recruited into microbes-associated with edible parts of plants. While human activities hasten the spread of AMR due to misuse and overuse of antimicrobials, it is critical to distinguish between naturally occurring- and newly acquired ARGs via anthropogenic process.

Apples are one of the most widely grown fruit crop in temperate regions and one of the most popular edible fruits in the world. Golden Reinders and Mandy apple cultivars are grown extensively in all apple-growing regions in Europe and, thus, have major economic value. Despite the health benefits and economic value of apples, it has been revealed that apples harbor diverse microbiomes [33–36]. Considering that plants are inherently colonized by microbes from the soil and adjacent niche, it is plausible that the bacteria harboring environmental resistome can be taken up by apple plants, and subsequently transferred to fruits, particularly when appropriate measures are overlooked. The processes, environmental pressures and pathways for the acquisition and transmission of these ARGs may further vary due to their unique ecological niches. Several studies have indicated that geographical location and genotypes affect the microbiome of apple [34, 37], while during post-harvest period, significant alterations can occur in the resistome composition [20]. However, the extent to which these factors (location-cultivar combinations) and microbial lifestyle interact and contribute to resistome variation remains unknown.

Here, we report the analysis of antibiotic resistance profiles and resistome of the bacterial species isolated from Golden Reinders and Mandy apples via culture-dependent technique and whole genome sequencing. In this study, we aimed to answer the following questions: [1] What are the effects of orchards-cultivar combinations, and microbial lifestyle (endophytic vs. epiphytic) on the antibiotic resistance profiles [2]? What are the association-patterns between bacterial hosts and ARGs, and how do these patterns relate to their ecological niche within the plant host? Overall, our study provides insights into the dynamics and prevalence of antibiotic resistance in agricultural environments. This is crucial for understanding the potential impact on crop health and the surrounding environment.

Materials and methods

Sampling and experimental design

Golden Reinders and Mandy apples were taken from four separate orchards (Fig. 1). A total of 96 apples per cultivar were collected from four orchards: IRTA Experimental Station in Mollerussa (EEL-Lleida) (41.617721, 0.870495), Esterri d’Aneu (42.624693, 1.122405), Fruits de Ponent in Alcarràs (41.5918610, 0.5631240) and Gotarta (42.421658, 0.764648). In every orchard, eight trees were selected at random and in each tree, six fruits were harvested using gloves. All fruit samples were visually examined to ensure uniformity in color, shape, quality, freshness, size and hardness. The collected apples were placed in sterile bags and transported immediately to the laboratory for further analysis.

Fig. 1.

Map displaying locations of the apple sampling sites (Map data ©d-maps.com). Golden reinders (indicated in green) were sampled from Gotarta (42.421658, 0.764648) and EEL-Lleida (41.617721, 0.870495) while Mandy apples (shown in red) were collected from Esterri d’Aneu (42.624693, 1.122405) and Fruits de Ponent (41.5918610, 0.5631240). Bacterial isolates from both epiphytic (surface) and endophytic (from surface-sterilized peels) of the apples were isolated and analyzed for antibiotic resistance

Cultivation

For the epiphytic bacteria, apple fruits were washed twice with washing solution (Tris-EDTA buffer with 2% Tween 80, pH 8.0). Each fruit was placed in a sterile washing bag containing 250 mL of the washing solution and shaken on a rotary shaker (Rotabit, J.P. Selecta, Barcelona, Spain) at 180 rpm for 20 min at room temperature. The washing solution from the two washes of each replica was pooled and centrifuged at 4000 × g for 20 min at 4 °C (Thermo Scientific, Grand Island, NY, USA). The resultant microbial pellet was resuspended in 20% glycerol and kept at − 80 °C. For the endophytic bacteria, the apples used for the recovery of epiphytic bacteria were thoroughly cleaned, surface-sterilized with 70% ethanol for 5 min and later immersed in 1% sodium hypochlorite solution for 5 min. Surface sterilization was then completed by rinsing the fruits two times with sterile distilled water. The final eluent from each fruit sample was plated on a Luria Broth agar (LB) medium and incubated at 25 ℃ for 24 h to verify sterilization. The apple fruits were peeled aseptically and placed in Stomacher bags (two peels per bag) containing 200 mL of Tris-EDTA + 2% Tween 80. For 90 s, the peels were mixed at maximum paddle speed (IUL Masticator Basic 470, Barcelona, Spain). The resulting solutions were centrifuged for 20 min at 4 °C, 4000 xg (Thermo Scientific, Grand Island, NY, USA). The resultant microbial pellets were resuspended in 20% glycerol and kept at − 80 °C (NuAire Blizzard Ultralow Freezer, Nirco, Madrid, Spain).

A total of seven different media were used for the isolation i.e., Minimal Culture Medium supplemented with different pigments (Chlorophyll, β-carotene and Quercitin) and nitrogen sources (Alanine, Glycine and Aspartic Acid) [38]. A rich medium, LB agar was used for growth control. The pigments and nitrogen sources were added into the medium to mimic the physiological, biochemical and nutrient composition in apple fruits [39, 40]. Minimal Culture Medium was prepared as follows per litre of water: 15 g agar, 19 mL 50X salt concentration solution (per litre of water: 26 g KCl, 26 g MgSO₄·7H₂O and 76 g KH₂PO₄) and supplemented with the nitrogen and pigment sources as described in Supplementary Table 1. Cycloheximide (100 ppm) was added into the media inhibit fungal growth. During the isolation process, frozen pellets were serial-diluted to 10-folds and 100 µL from dilutions: D1 (10⁻¹), D2 (10⁻²) and D3 (10⁻³) for epiphytic and D0 (10 ⁰), D1 (10⁻¹) and D2 (10⁻²) for endophytic samples were subsequently plated onto the media. The plates were incubated at 25 °C for 48–96 h. After incubation, single bacterial colonies with different morphologies were picked and sub-cultured on new LB plates to purify the bacterial isolates. Pure cultures were preserved at − 80 °C in glycerol. Pure isolates (both epiphytes and endophytes) were sent to the Instrumental Techniques Laboratory (University of Leon) for the identification.

Molecular analysis

Bacterial cells were mechanically disrupted by bead beating (6 ms⁻¹, 20 s). The lysates were incubated at 95 °C for 10 min and then centrifuged for 5 min at 5000 rpm. The 16 S ribosomal RNA genes of all isolates were amplified using 2 µL of the supernatant and the universal bacterial primer pair 27 F (5′AGAGTTTGATCMTGGCTCAG) and 1492R (5′TACGGY TACCTTGTTACGACTT) [41], 0.5 µM each primer, in a 50 µL PCR reaction with 1x Taq-&GO Ready-to-use PCR Mix (MP Biomedicals, Eschwege, Germany). The reaction mixtures were subjected to the following reaction conditions repeated for 25 cycles: initial activation of enzyme for 4 min at 98 °C, denaturing at 98 °C for 30 s, primer annealing at 48 °C for 30 s, 90 s for chain extension at 72 °C followed by final extension step at 72 °C for 5 min and lastly holding at 4 °C infinite. The 1400-bp DNA fragments were purified (Wizard^® SV Gel and PCR Clean-Up System, Promega, Walldorf, Germany) and Sanger sequenced using a bidirectional sequencing approach.

Phenotypic antibiotic resistance screening

The Kirby-Bauer disc diffusion method was used to assess antibiotic resistance in compliance with the guidelines provided by the European Committee of Antimicrobial Susceptibility Testing (EUCAST). The quality control reference strains were not included due to limited access of certified cultures; however, all assays were performed under controlled and standardized laboratory conditions. The discs were impregnated with seven standard antibiotics belonging to different generations as follows: Tetracyclines (tetracycline- 20 µg/mL), β-lactams (penicillin G- 10 µg/mL), quinolones (ciprofloxacin- 5 µg/mL), cephalosporins (ceftriaxone- 30 µg/mL), glycopeptides (vancomycin- 50 µg/mL), carbapenems (imipenem- 10 µg/mL) and macrolides (erythromycin- 30 µg/mL). The disc diffusion assays were carried out in the following manner: [1] Müller-Hinton agar was streaked with bacterial samples whose turbidity corresponded to 0.5McFarland [2], the discs impregnated with antibiotics were placed on the media and [3] incubated at 30 °C for 24 h to ensure consistent and optimal growth. Extra care was taken during screening to adhere to the 3 × 15 min rule, which states that none of the three steps should take longer than 15 min. Finally, the zone of inhibitions were measured using a meter-rule and recorded. EUCAST’s epidemiological cutoff (ECOFF) values were utilized to determine antibiotic resistance. Clinical breakpoints of EUCAST were applied to antibiotics that lacked a defined ECOFF value. Breakpoints from the Clinical and Laboratory Standards Institute (CLSI) were used in cases where EUCAST recommendations did not establish any value. The isolates that demonstrated resistance to atleast three distinct antibiotic classes were classified as multidrug-resistant and were further considered for whole genome sequencing.

DNA extraction and whole genome sequencing (WGS)

We selected 18 multidrug-resistant isolates for the next step (whole genome sequencing) according to the multi-criteria scheme i.e., isolates were chosen to maximize taxonomic breadth and phenotypic diversity (isolates with distinct resistance patterns across the tested antibiotics), as well as metadata representation of both epiphytic and endophytic isolates from multiple orchards. DNA from the selected isolates was extracted using FastDNA™ Spin Kit. The genomic DNA was quantified using Nano drop (IMPLEN, NP80). The whole genome of 18 multidrug isolates were sequenced by Novogene (UK). Libraries were prepared using Illumina-compatible library preparation reagents according to the provider’s standard protocols and sequenced using Illumina NovaSeq 6000 (platform, PE150).

Bioinformatic analysis

The quality of the raw reads were assessed using FastQC [42] and filtered using Trimmomatic v0.36 [43] to remove adapter sequences and low-quality bases. SPAdes (v1.2.10) at default parameters was used for de novo assembly of the reads into contigs, which were subsequently binned into genomes assemblies [44]. The preliminary assemblies were polished to remove errors like base-pair mismatches and indels using pilon [45]. The contiguity and completeness of both the SPAdes and pilon assemblies were then evaluated using QUAST [46]. The quality of genomes was calculated using CheckM [47] by estimating lineage-specific marker genes. Genomes that failed one or more quality control were excluded and only 10 genomes passed downstream quality-control thresholds (≥ 90% completeness and ≤ 5% contamination; however, genomes with slightly exceeding the contamination threshold but with high completeness, were retained to ensure inclusion of valuable genomic information) (Supplementary: Additional file 1). To generate draft genomes, contigs were re-ordered against reference genomes obtained from NCBI (ncbi.nlm.nih.gov/genome) using ragtag [48] with the default alignment parameters. PlasmidSPAdes v3.12 [44] was used to assemble plasmid-contigs from the genome assemblies. Using a genome-masking database of all closed chromosomes, MOB-suite v. 3.0.0 [49] was employed to predict, reconstruct and classify plasmids from the draft contigs assemblies according to default parameters (≥ 80% coverage and identity sequence). The potentially complete plasmid contigs were identified from the MOB-recon output by selecting contigs flagged as circular and classified as a plasmid (i.e. those that passed the chromosome-masking phase). To detect mobile genetic elements (MGEs), we used Skandiver [50]. AMR genes annotations of both plasmids and chromosomes were identified in chromosome, plasmid and other MGEs sequences using ABRicate tool (version 0.8.13), with default settings of 90% nucleotide similarity and coverage alignment above 50% [51]. The following databases were used for AMR annotation: Antibiotic Resistance Gene-ANNOTation (ARG-ANNOT) [52], Comprehensive Antibiotic Resistance Database (CARD) [53], ResFinder [54], , MEGARes [55], National Center for Biotechnology Information (NCBI) [56], AMRFinderPlus [57] and PlasmidFinder [58]. To ensure consistency across the databases, redundant ARG hits from different databases were merged based on the gene name and sequence similarity (≥ 90% nucleotide identity and ≥ 80% coverage). The software Prodigal version 2.6.3 [59] was utilized to predict open reading frames. We defined if mobile ARG is located 5000 base pairs upstream and downstream from the detected MGE [60].

Statistical analyses

Phenotypic resistance profiles of the isolates were determined using antibiotic susceptibility testing against a range of antibiotics, including tetracyclines (TET), quinolones (QUI), cephalosporins (CEP), β-lactams (LAC), macrolides (MAC), glycopeptides (GLY) and carbapenems (CAR). Venn diagram was generated using an online tool (http://www.ehbio.com/test/venn/#/) to visualize the overall distribution of antibiotic-resistant isolates. To ascertain the impact of orchard-cultivar combinations and microbial lifestyle (epiphytes vs. endophytes) on the antibiotic resistance profiles, we conducted a multivariate statistical analysis using vegan package in R v3.6.1 [61]. Prior to PERMANOVA, we tested the assumption of homogeneity of multivariate dispersions using the betadisper () function in the vegan package [62] and no significant differences were detected (P > 0.05) (Supplementary file: Table 2). We calculated Bray-Curtis dissimilarity distance, which quantified the differences in resistance patterns among isolates. Subsequently, we performed a PERMANOVA (adonis2 function) to determine whether orchard-cultivar combinations or microbial lifestyle had significant effects on the resistance profiles. Hierarchical clustering was conducted using Ward’s method [63] to the Bray-Curtis dissimilarity matrix of the phenotypic resistance to visualize resistance patterns. Since our data were binary (presence/absence of resistance), no additional standardization was done, as Bray-Curtis dissimilarity metric used to compute pairwise distances inherently accounts for differences in resistance patterns across the isolates. To investigate the taxonomic distribution of the resistant isolates, we constructed a phenotypic resistance-based phylogenetic tree and validated the taxonomic assignments using GTDB-Tk [64]. The phylogenetic tree was generated by clustering isolates according to their antibiotic resistance profiles. The generated tree was visualized using iTOL v6 [65], which included multiple annotation layers to highlight distinct isolates characteristics. To further examine the resistance profiles of the isolates, we calculated the proportion of resistant isolates per antibiotic class.

We classified ARGs according to their genomic localization (chromosomal, plasmid and other MGEs). The number and proportion of ARGs in each category were calculated, and the relative abundance of ARGs was plotted using ggplot2 in R version 4.4.1 [66]. The presence of ARGs conferring resistance to specific antibiotic classes was visualized using an UpSet plot [67], which highlights shared and unique resistance patterns among isolates. We employed pheatmap package [68] to generated heatmap illustrating the presence, localization and antibiotic classification of ARGs. The percentage of ARGs conferring resistance to antibiotic classes, was visualized using a pie chart. To further analyze the relationship between bacterial hosts and their ARGs, the network analysis was computed using Cytoscape version 3.6.1 [69]. Canonical Correspondence Analysis (CCA) was performed in R version 4.4.1 using the vegan package [70]. The analysis included the response matrix (resistance phenotypes) and the explanatory matrix (presence/absence of ARGs). The data were Hellinger-transformed and the significance of the canonical axes was assessed using 999 permutations (anova.cca function).

Results

The effect of orchard-cultivar combinations and microbial lifestyle on the antibiotic resistance profiles

A total 516 bacterial, comprising of 306 epiphytes (59.3%) and 210 endophytes (40.7%) were subjected to antibiotic susceptibility screening. From these, 272 isolates were found to be resistant to at least one of the seven antibiotics tested, from which 203 were classified as epiphytes and 95 as endophytes. Among those, 26 overlapped between the two groups (Fig. 2A). Eighty isolates showed resistance to more than three different classes of antibiotics. The observed resistance profiles varied across the antibiotics, with over 50% of the isolates exhibiting resistance to tetracycline (69%), quinolones (represented by ciprofloxacin, 61.6%) and cephalosporins (represented by ceftriaxone, 57.9%). In contrast, none of the isolates showed resistance to carbapenems (represented by imipenem) (Fig. 2B and Supplementary file: Fig. 1). We found that microbial lifestyle (epiphytes vs. endophytes), was the main determinant of the antibiotic resistance profile (R² = 0.55411, P = 0.028*, Supplementary file: Table 3). The hierarchical clustering distinctly separated epiphytes and endophytes into two clusters, underscoring the significant influence of microbial lifestyle on the resistance profiles (Fig. 2C). Epiphytes were resistant to a higher number of antibiotics (with proportion of 63.6%) compared to endophytes (35.3%), with shared resistance across tetracyclines, cephalosporins and quinolones.

Fig. 2.

Comparative assessment of antibiotic resistance profiles. A Venn diagram illustrating overall count of the strains resistant to specified antibiotics. B Percentage of the resistant isolates to antibiotics (TET-tetracycline, QUI-quinolone, CEP-cephalosporin, LAC-lactam, MAC-macrolide, GLY-glycopeptide and CAR-carbapenem) in each sampling point (EEL-Lleida, Gotarta, Esterri and Fruits de Ponent (FdP)). Golden Reinders were collected from EEL-Lleida and Gotarta while Mandy apples were collected from Esterri and Fruits de Ponent (FdP). Different colors provided in the panel represent sampling sites. C Comparison of resistance profiles between epiphytic and endophytic isolates from four sampling sites. The hierarchical clustering was based on the Bray-Curtis dissimilarity matrix derived from their respective resistance profiles. Figure 1: Map displaying locations of the sampling sites (Map data ©d-maps.com)

Orchard-cultivar combinations (with R² = 0.21138, P = 0.897) had no significant effect on the observed antibiotic resistance (Supplementary file: Table 3). Esterri-Mandy exhibited the most distinct resistance profiles, particularly in tetracyclines (with proportion of 49.1%), quinolones (42.0%) and cephalosporins (39.9%) compared to other orchards. The isolates from EEL-Lleida- and Gotarta-Golden Reinders showed higher resistance to tetracyclines (34.8%), lactams (19.4%), macrolides (17.4%) and glycopeptides (15.1%) while the isolates from Esterri- and FdP-Mandy exhibited higher resistance to cephalosporins (33.5%) and quinolones (33.5%) (Fig. 2B).

The drivers of the antibiotic resistance profiles

The taxonomic profile of resistant isolates varied across microbial lifestyle (epiphytes vs. endophytes). The predominant phyla associated with these isolates were Actinomycetota, Proteobacteria and Bacillota (Supplementary file: Fig. 2). We observed that Micrococcales, Bacillales and Enterobacterales were the most dominant orders (Fig. 3A). Notably, Micrococcales were predominant across the microbial lifestyle and exhibited a significant resistance to cephalosporins (with proportion of 62.2%) and quinolones (61.9%). Additionally, Enterobacterales were frequently identified in epiphytes but seldom in endophytes (Fig. 3B and C). These bacteria exhibited multiple resistances to glycopeptides (69.8%), macrolides (60%), lactams (52.3%) and tetracyclines (32.8%).

Fig. 3.

Taxonomic profiles of the resistant isolates. A Phylogenetic tree showing the taxonomical classification of the bacterial orders according to their antibiotic resistance profiles. Different colors in R1 indicates bacterial taxonomy, R2 shows the microbe lifestyle (blue-epiphytes and green-endophytes) and R3 represent the antibiotic resistance profiles of the isolates (white, grey, and black indicates susceptible, intermediate, and resistant phenotypes, respectively). B and C More detailed classification of the data shown in A, wherein bacterial resistance profiles are depicted in a distinct panel (Epiphyte vs. Endophyte). The y-axis of each panel denotes the percentage and the number of resistant isolates, while the x-axis displays the antibiotics (TET (Tetracyclines), LAC (lactams), CEP (cephalosporins), QUI (quinolones), MAC (macrolides) and GLY (glycopeptides)). Stacked bars illustrate bacterial orders conferring resistance to the antibiotic classes, with a color-coded legend for the bacterial orders provided in the panel

The structure and localization of the ARGs in bacterial genomes

From the 10 genomes that passed downstream quality-control thresholds, we observed that the distribution of the ARGs varied across the microbial lifestyle (epiphytes vs. endophytes). Majority of epiphytic bacteria had higher number of ARGs compared to endophytes. Conversely, endophytic bacteria harbored exclusively chromosomally related ARGs (Fig. 4A). In both microbial lifestyles, the proportion of ARGs located in chromosome was higher than those in MGEs. Nevertheless, we solely observed plasmid-encoded ARGs in Staphylococcus arlettae. The ARGs were categorized into 14 antibiotic classes (Fig. 4B). The most frequently detected ARGs were associated with multi-drug resistance (with proportion of 33%), followed by aminocoumarin (13%) and rifamycin (12%) (Fig. 4B–D). In contrast, lincosamide and diaminopyrimidine (both with 1%) resistance was observed at the lowest frequency. Besides multi-drug resistance, only a few target-specific ARGs were annotated, including (Rif) iri for rifampicin resistance in Rhodococcus qingshengii, and lnu(A) and FosD in Staphylococcus arlettae conferring resistance to lincomycin and fosfomycin respectively. Pantoea agglomerans (with 26 ARGs) was significantly observed to harbour higher diversity of ARGs, while Rhizobium pusense and Agrobacterium pusense had a lower overall abundance, with only two ARGs (Afab_ACT_CHL and ceoB). Mobile genetic elements such as plasmids are known as reservoir and vectors for the ARGs dissemination. It was noted that Streptomyces species and Staphylococcus arlettae harbored higher number of MGE-associated ARGs than other bacterial isolates. Interestingly, we found two genes including mtrA and rpoB on either chromosomes or MGEs across the bacteria.

Fig. 4.

Distribution of ARGs in bacterial isolates. A The number and possible localization of ARGs within bacterial groups. The blue, green and orange colors indicate chromosome, other MGEs and plasmid, respectively. B Proportion of antibiotic classes to which the ARGs confer resistance. C Heatmap showing distribution and localization of ARGs within the bacterial isolates as well as the drug classes to which they confer resistance. D The upset plot depicts the intersection of bacterial isolates with the antibiotics classes to which they are resistant

The bacterial host-ARG associations

To further explore the relationship between microbial lifestyle, isolates and ARGs, network analysis between bacterial hosts and ARGs was constructed. The network in epiphytes comprised of 43 nodes (6 bacterial species and 48 ARGs) with 46 edges whereas the network in endophytes consisted of 36 nodes (4 bacterial species and 34 ARGs) with 34 edges (Fig. 5A). Eight bacterial isolates were identified as potential hosts for the ARGs. Two Streptomyces species (Streptomyces californicus and Streptomyces microflavus) had strong association with multiple ARGs. We observed that two ARGs (ceoB and Afab_ACT_CHL) associated with the same endophytic bacteria i.e. Agrobacterium pusense and Rhizobium pusense. On the other hand, eight ARGs (oleC, novA, tet (43), otrC, gimA, r39_beta-lactamase, soxR and cmlv) were associated exclusively within Streptomyces species. Resistance gene rpoB2 was observed in Streptomyces species and Rhodococcus qingshengii from the same epiphytic group (Fig. 5A) while mtrA, parY and rpoB were associated with distinct bacterial isolates from different bacterial groups. Two bacteria, namely Pantoea agglomerans and Staphylococcus arlettae possessed distinct ARGs that did not overlap with any group. Notably, majority of the ARGs were mainly annotated to MDR which confer resistant to a wide-range of antibiotics and toxic compounds.

Fig. 5.

The bacterial host-ARG associations. A Network analysis displaying the associations between bacterial isolates and the ARGs. The nodes were colored according to bacterial groups or ARGs (blue, green and pink colors denote endophytes, epiphytes and ARGs respectively) and the size is proportional to the number of connections (degree). The edges indicate the observed host-ARG associations and the width correspond to edge betweenness value. B Canonical correspondence analysis (CCA) of the relationship between explanatory variables (ARGs), response variables (resistance phenotypes) and the bacterial isolates. Black arrows and red arrows represent ARGs and resistance phenotypes respectively. The bacterial isolates are named in blue

Canonical correspondence analysis showed that the explanatory variables accounted for 87.4% of the total variation (Fig. 5B, Supplementary file: Table 4). Resistance to glycopeptide (GLY), quinolone (QUI) and lactam (LAC) were clearly separated from the other resistance phenotypes. The main contributors to the resistance phenotypes were different, we observed that ARGs (msrA and mphO) were strongly associated with macrolides (MAC) and quinolones (QUI) respectively. Viomycin_phosphotransferase and cmlv had weak effect on the ARGs contribution, but were significantly correlated to lactam phenotypes. Moreover, tetracyclines (TET9), cephalosporins (CEP) and Macrolides (MAC) phenotypes showed a moderate interrelationships and influence from multiple ARGs.

Discussion

In the present study, we identified the effects of orchard-cultivar combinations and microbial lifestyle (endophytes vs. epiphytes) on the antibiotic resistance profiles. Our results show that 1) majority of the bacteria demonstrated notable resistance to tetracycline followed by quinolones, cephalosporins and lactams, while none of the isolates were resistant to carbapenems [2], microbe lifestyle, i.e., epiphytes vs. endophytes, was identified as the main determinant of the observed antibiotic resistance profiles, highlighting distinct microbial community composition in each compartment inherently carry different ARGs repertoires [3], Micrococcales, Bacillales and Enterobacterales were recognized as the main key drivers that shape the antibiotic resistance profiles [4], majority of ARGs encoded for multidrug (MDR) resistance, however only a few target-specific ARGs were annotated, including (Rif) iri (rifampicin), lnu(A) (lincomycin) and FosD (fosfomycin) and [5] Staphylococcus arlettae was shown to possess more plasmid-encoded ARGs, suggesting its role to participate in HGT.

In our findings, most of the isolates exhibited resistant to tetracycline. This may be due to long term usage of tetracycline over the past decades, driven with their high efficiency, affordability and broad-spectrum activity, which may have favored the selection of tetracycline resistance genes in the environment, culminating in heightened phenotypic resistance [71–74]. More than 37 genes confer resistance to tetracyclines either cooperatively or independently [71]. The most common mechanism is tetracycline efflux, which often acquired through HGT as they are linked to MGEs with broad host range [75]. Moreover, their co-inheritance with other resistance genes in the same MGEs, further hasten the spread of tetracycline resistance in the environment [76, 77]. Unlike tetracyclines, which are extensively used in agriculture, carbapenems are rarely employed outside clinical settings. This reduced exposure suggests that plant-associated bacteria are less prone to evolve resistance mechanisms against it. Notably, these resistance patterns we observed were strongly influenced by microbial lifestyle with clear differences between epiphytes and endophytes. This is indicative of an important role which the ecological niches of microorganisms play in the shaping their adaptability [78–80]. In the fruit episphere, myriad factors such as microbial interaction, different environmental conditions, pesticide treatments and antimicrobial residues may contribute towards the increased pool of resistant bacteria [81–83]. This ever-changing environment might lead to the establishment of microbial population harboring a whole plethora of genetic adaptations including acquisition of ARGs. Orchard-cultivar specific variations in antibiotic resistance profiles were detected, implying that management practices, and environmental factors influence bacterial resistance patterns. Previous studies have demonstrated that the use of agrochemicals and farming practices affects the composition and functionality of resistome in plant-associated bacteria [84–86]. Following the indications that the Enterobacterales prefer epiphytic niches, it can be hypothesized that they have evolved strategies for the adaptation and selection of its resistant phenotypes under various environmental stress [10, 87].

Whole genome analysis of the sequenced isolates revealed that the observed resistance profiles were determined at genetic level. In theory, resistance genes can be potentially acquired from any sources [9], thus the antibiotic resistome from the two apple cultivars might be shaped by several factors. In the farm, ARGs contamination can arise from irrigation water, contaminated raw sewage, wild animals and direct usage of antimicrobials, particularly in animal husbandry or waste water treatment plants which have been shown to systematically co-select MGEs carrying multiple ARGs [11, 88]. The selection pressure exerted by the aforementioned factors, together with the possible introduction of microbes from non-agricultural sources, may result in dysbiosis by altering the intrinsic resistome and microbial community composition. Even in the absence of antibiotics, those factors may have pleiotropic impacts on different phenotypes, including the concurrent emergence of resistance [89]. This suggests that fruit episphere may be more exposed to external sources, which may reshape the resistance genes. The higher chromosomal ARGs in isolates further underlines potentially stable intrinsic resistance mechanisms within bacterial communities, as previously demonstrated in studies of apples microbiome and resistome in the post-harvest period [20] and an exemplary comparison of sphagnum moss resistome in undomesticated bog system [8]. The identified ARGs conferred resistance to 14 different drug classes, with the most prevalent being peptides, aminoglycosides, tetracyclines, beta-lactams and aminocoumarin. The fact that each of these classes of antibiotics targets specific bacterial processes, their resistance mechanisms are varied and well established. Previous research have identified ARG-types, which are cooperatively or independently capable of conferring resistance to tetracycline, β-lactams, aminoglycosides, methicillin, macrolides and sulphonamides [90–92]. These ARGs were grouped based on their biochemical pathways used to provide resistance, including modifying drug targets, reducing antibiotics influx, exporting antibiotics via efflux pumps or inactivating antibiotics with specific enzymes [91]. Interestingly, most of these drug types are (semi-) synthetics and their widespread presence aligns with extensive usage in both clinical and agricultural settings [93], which likely drives the selection of ARGs in bacterial communities. In contrast, the unique mechanisms of action, the restricted clinical use and limited gene mobility might account the lower prevalence of diaminopyrimidine and nitroimidazole resistance. Historically, Pantoea agglomerans has been acknowledged as plant promoter, biocontrol agent, seed colonizers and opportunistic pathogen that causes plant diseases such as soft rot [94, 95]. Furthermore, it is known to cause infection in immunocompromised patients [96]. This double-edged threat of antibiotic resistance i.e., having higher number of ARGs and pathogenicity bears serious implications for agriculture, as managing these bacteria may become more difficult over time. However, it needs a strain-specific evaluation and analysis in context of the whole plant microbiome [97].

The primary cause of ARGs transfer is widely acknowledged as HGT mediated by MGEs. ARGs linked with MGEs are even more “global” than microorganisms, as illustrated by the capacity of mobile ARGs to span habitat boundaries [98]. The annotated plasmid encoded-ARGs in Staphylococcus arlettae presented here suggests its significant role in ARGs dissemination via HGT within fruit microbiomes. Besides plasmid, we showed that other MGEs such as transposons, integrons and insertion sequences (IS) were also important ARGs carriers. We observed certain resistance genes on either chromosomes or MGEs. This suggests the possibility of both vertical inheritance and horizontal gene transfer (HGT) among plant microbiomes. These interactions not only strengthen the resilience of microbial communities but also raise serious concerns about the spread of resistance characteristics to pathogenic species via food chain [9, 99]. The ecological niches, selective and the accessibility of MGEs may have all play a role in the observed heterogeneity of ARGs diversity across bacterial species [100].

The network analysis exhibited significant disparities between epiphytic and endophytic bacterial communities. We discovered that epiphytic bacteria had greater number of ARGs than endophytes. Streptomyces species, particularly Streptomyces microflavus and Streptomyces californicus, were significant contributors to the resistome, associating with multiple ARGs. These associations imply that Streptomyces species might be critical ARGs reservoirs and distributors to diverse bacterial communities. Streptomyces species are well known for their ability to produce antibiotics [101, 102] and their possession of multiple ARGs indicates that they may have adapted to living in environments in which antibiotic production is a common strategy for competing with other microbes or preventing self-inhibition [103, 104]. Resistance genes such as Afab_ACT_CHL and ceoB were endophyte-specific. Fruit endosphere have been shown to promote microbe-interactions and facilitate the exchange of genetic material including ARGs [22, 23]. Studies have demonstrated that endophytic bacteria experience greater exposure to plant internal chemical milieu [105], perhaps leading to evolution of resistance characteristics. Interestingly, ARGs that did not overlap with other groups were observed in Pantoea agglomerans and Staphylococcus arlettae. This finding is noteworthy because it raises the possibility that some bacterial species have unique or uncommon ARGs that have not yet proliferated across other bacterial population. We discovered that specific variables significantly influenced the association between ARGs and observed resistance phenotypes. Resistance phenotypes of lactams (LAC), quinolones (QUI) and glycopeptides (GLY) were clearly distinguishable from other phenotypes. ARGs such as mphO and msrA, demonstrated strong correlation with quinolone (QUI) and macrolide (MAC) phenotypes, respectively. On the other hand, tetracyclines, macrolides (MAC) and cephalosporins (CEP) exhibited moderate interrelationships, suggesting that multiple ARGs influence these phenotype traits. The prevalence of multidrug resistance (MDR)-related ARGs further emphasizes the adaptability of these bacterial communities to withstand a wide range of antimicrobial stressors.

Conclusion

The study highlighted that orchard-cultivar combinations and microbial lifestyle are two variables responsible for the observed antibiotic resistance profiles. These variables reflect the environmental exposure, micro-ecological adaptation and host tissue selectivity of functional resistance. The results for hierarchical clustering showed not only that epiphytic bacteria exhibited higher resistance profiles but also wider diversity of ARGs than endophytes, expressing unique resistance phenotype. Micrococcales, Bacillales and Enterobacterales are the most predominant orders based on taxonomic analysis, underlining their prominent roles in shaping the resistance patterns. Although WGS data plays a crucial role in revealing the genomic patterns consistent with the observed functional profiles, these findings should be viewed as preliminary. They provide an initial exploration of the diversity of ARGs, which may not give a complete characterization of the resistome due to the limited number of genomes analyzed. Overall, the findings of the study underscored the interplay between orchard environment, cultivar genotypes, and microbial lifestyle in shaping antimicrobial resistance, carrying implications for AMR management in agricultural settings.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

PCR

Polymerase chain reaction

DNA

Deoxyribonucleic acid

rRNA

Ribosomal ribonucleic acid

MGEs

Mobile genetic elements

ARGs

Antibiotic resistance genes

LB

Luria Broth agar

EDTA

Ethylenediaminetetraacetic acid

EUCAST

European committee of antimicrobial susceptibility testing

Author contributions

D.K., A.M.S., N.T., R.M., G.B., and A.A. conceptualized the study. D.K., D.K.R., W.A.W., and A.A. performed data analysis. D.K., D.K.R., W.A.W., and A.A. interpreted the results. D.K. wrote the first manuscript draft. All authors read, reviewed, and approved the final manuscript.

Funding

This work was supported by the scholarship fund from the German Academic Exchange Service (DAAD) and the University of Lleida and IRTA predoctoral UdL-IRTA Sponsored Fellowship 2021. Part of this work has been financially supported by the Spanish ‘Agencia Estatal de Investigación’ (AEI) and the European Regional Development Fund (ERDF) through the national project PID2020-117607RR-I00 (ENVIRONAPPLE). Additionally, this work has been supported by the 2021 SGR 01477 grant and the CERCA Programme from the ‘Generalitat de Catalunya’.

Data availability

The Antibiotic Susceptibility Screening Dataset has been newly generated. These data are openly available in Figshare at 10.6084/m9.figshare.29617145. The Culture Media Composition Dataset is newly generated and is detailed in Supplementary file: Table 1. The test for Homogeneity of Multivariate Dispersion using Betadisper Result Dataset is newly generated and is referenced in Supplementary file: Table 2. The Statistical Analysis and PERMANOVA Results Dataset is newly generated and can be found in Supplementary file: Table 3. The Canonical Correspondence Analysis (CCA) Eigenvalues and Explained Variation Result Dataset is newly generated and is referenced in Supplementary file: Table 4. The Associations between Bacterial Hosts and ARGs Dataset is newly generated and is referenced in Supplementary file: Table 5. Genomes Contamination Check Dataset is newly generated and is referenced in Supplementary: Additional file 1. The 16 S rRNA Gene Sequencing Identification are available in SRA NCBI repository via PRJNA1265908 and can be accessed through the link: https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA1265908. The WGS data have been deposited in the [SRA NCBI] repository (Accession number: PRJNA1243104) and can be accessed from the following link: https://dataview.ncbi.nlm.nih.gov/object/PRJNA1243104?reviewer=6v0gfahlu0tujp84qehbsl2mvd.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.