The spread of Acinetobacter species, especially Acinetobacter baumannii, has been associated with an alarming increase in hospital mortality rates, mainly affecting immunocompromised patients in intensive care units (ICUs) [38]. However, A. baumannii is not the only clinically relevant member of the genus. In recent years, several non-baumannii species have gained medical importance due to their increasing ability to cause nosocomial infections. Among them are Acinetobacter haemolyticus, A. lwoffii, A. ursingii, A. parvus, and A. junii, which are part of the Acinetobacter calcoaceticus-baumannii complex [26].

The emergence of infections caused by these species has been favored by the presence of multiple antimicrobial resistance mechanisms in Acinetobacter spp. Carbapenems, considered first-line drugs due to their broad spectrum of activity and high efficacy against multidrug-resistant bacteria, have lost effectiveness as resistant strains have emerged [39]. This phenomenon is not limited to A. baumannii (CRAB) but has also been documented in A. haemolyticus, significantly complicating the treatment of infections caused by this pathogen [40]. In Mexico, the isolation of multiple clinical strains of A. haemolyticus has been reported, along with the presence of resistance genes such as blaNDM−1, evidencing its emerging role as a potential reservoir and disseminator of antimicrobial resistance [26, 41].

Antibiotic resistance monitoring in the Acinetobacter genus has been biased by the almost exclusive reporting of A. baumannii strains due to its higher frequency of isolation, leading to the overlook of acquisition of relevant resistance genes, such as blaNDM−1, in non-baumannii species like A. haemolyticus [26, 28]. This phenomenon affects epidemiology and clinical practice, complicating treatment decisions, especially now that new mechanisms of antibiotic resistance dissemination are being reported [42, 43]. Studies on releasing outer membrane vesicles (OMVs) have shown that resistance gene dissemination no longer relies solely on the previously studied conventional mechanisms [44]. In this work, we analyzed the role of OMVs in carbapenem-resistant A. haemolyticus under antibiotic-induced stress [44]. A. haemolyticus AN54 is a clinical strain that harbors resistance genes such as blaOXA−265, aac(6’)-Ig, aphA6, and blaNDM−1, present in one of the five plasmids of the strain, as reported in a previous study [26].

The release of resistance genes via OMVs has been reported in P. aeruginosa, K. pneumoniae, E. coli, and A. baumannii, where genes are transferred between bacteria of the same species and genus and different genera [2, 15,16,17]. However, the number of studies regarding active beta-lactamases released via OMVs has been limited.

It is important to note that the intrinsic protein characteristics of beta-lactamases are key to their release via OMVs [33]. NDM-1, a metallo-beta-lactamase with unique properties, has rapidly spread since its first identification in 2009 [45]. Its anchoring to the membrane, mediated by its signal peptide and residues such as Cys26, Arg52, and Arg45, protects it from degradation by periplasmic proteases under low Zn (II) conditions and favors its packaging and release in OMVs as an active enzyme capable of contributing to bacterial resistance [18, 30, 33, 43].

In our study, we tested different conditions, incubation times, and carbapenem concentrations to obtain OMVs from our A. haemolyticus strain carrying NDM-1; because it has been reported that longer incubation times or severe stress result in a greater probability of obtaining E-type OMVs (cell lysis) rather than B-type ones (natural secretion) [46]. Under these conditions, transmission electron microscopy (TEM) demonstrated that AN54 releases B-type OMVs without altering membrane morphology, even under antibiotic stress. This stability may be attributed to NDM-1, which is anchored to the membrane and does not compromise bacterial fitness, as previously shown by López et al. [30]. These observations contrast with previous reports in E. coli, where peptidoglycan-level membrane alterations have been associated with OXA-type beta-lactamase expression, suggesting that envelope effects may differ depending on the type of beta-lactamase carried by each bacterium [30, 33, 47].

Additionally, we observed no differences in phenotype between OMVs released by the strain AN54 (with blaNDM−1) and the control strain AN54Δe (without blaNDM−1) by TEM. However, we observed an increase in OMVs numbers when the bacteria were exposed to antibiotic stress, suggesting that antibiotic pressure favors increased OMVs release, likely as a protective mechanism against the effects of the antibiotic on the secreting bacteria. These results coincide with those of Kesavan et al. [21], who observed increased release of OMVs after eravacycline induction in A. baumannii ATCC 19,606 and A. baumannii JU0126. Nevertheless, they observed no significant phenotypic differences between antibiotic-induced OMVs and OMVs obtained without induction. On the other hand, although the antibiotic resistance phenotype of the strains has not been linked to OMV morphology, OMVs from antibiotic-resistant strains have been observed to be more cytotoxic and immunogenic [24].

When assessing carbapenemase activity in OMVs released by AN54 under noninduced antibiotic conditions, carbapenemase activity was detected. These results suggest that, in the absence of antibiotic stress, released OMVs can degrade antibiotics in the environment. However, when OMV secretion was induced under antibiotic stress, they were able to completely degrade carbapenems in the medium. These findings could have important clinical implications for infections caused by carbapenemase-producing bacteria, as in the case of our strain AN54, which secretes OMVs harboring active NDM-1 carbapenemase even before the presence of antibiotics. Our results also suggest that OMVs can disseminate active carbapenemases, even in the absence of the antibiotic, thereby contributing to the spread of resistance among bacterial populations; however, further studies are needed to confirm this potential mechanism.

Carbapenemase activity mediated by OMVs has been previously reported in different models, such as Stenotrophomonas maltophilia (after imipenem, amoxicillin, and ticarcillin exposure), in Klebsiella pneumoniae (showing KPC and NDM-1 activity with meropenem induction), in Escherichia coli (showing NDM-1 activity with meropenem induction), and in Acinetobacter baumannii (showing OXA-24 and OXA-58 activity) [7, 16, 34, 43, 48,8,]. This is the first study of an A. haemolyticus strain carrying NDM-1 in which OMV secretion was induced by two carbapenems at different concentrations, reveling greater activity against imipenem in the OMVs, attributable to NDM-1’s affinity for this antibiotic.

To determine whether the presence of the blaNDM−1-carrying plasmid and exposure to antibiotics alter vesicle contents, we analyzed the OMV protein profile obtained from AN54Δe and AN54 under different conditions. We observed variations in the protein profiles of the OMVs analyzed and identified common protein bands that were also conserved in the total bacterial extract, suggesting that some membrane proteins of the secretory strain are packaged into OMVs regardless of the stimulus. On the other hand, when analyzing the protein profiles of AN54 OMVs induced versus those of the AN54Δe control, we observe even greater variation. OMVs induced by imipenem exhibited a more defined profile, possibly due to more stable vesiculation and differences in protein synthesis. In contrast, induction with meropenem could elicit a more complex response, as reflected by greater differences in protein bands, although further studies are needed to corroborate this. However, our findings demonstrate that, despite belonging to the same antibiotic class, the two drugs differentially affect the protein profile of OMVs, underscoring the importance of performing proteomic analysis.

Proteomic analysis revealed that most of the proteins present in the OMVs come from the cytoplasm, outer membrane, and ribosomes, which is consistent with the findings of Dhurve et al. and Kesavan et al., who predominantly identified cytoplasmic proteins, followed by outer membrane and periplasmic proteins in OMVs derived from A. baumannii under antibiotic-free conditions and after exposure to eravacycline, respectively [21, 22].

These results are consistent with the theory proposed by Toyofuku et al., who suggested that type B OMVs may contain cytoplasmic proteins, a hypothesis supported by other studies [4, 21, 22, 49, 50]. The nature of OMVs can explain their protein content, serving as reservoirs of bioactive metabolites and cellular waste, which are beneficial to the bacterial community. It has been suggested that factors such as the anchoring of proteins at their final location and the involvement of multifunctional or “moonlighting” proteins could influence their packaging [11, 51, 52]. However, the exact mechanism of protein selection remains unclear, and cytoplasmic content in OMVs is still not fully understood, as the incorporation of periplasmic and outer-membrane proteins is more expected given their biogenesis. In the OMVs analyzed in this study, we identified proteins associated with essential metabolic pathways, including glycolysis, the Krebs cycle, and amino acid biosynthesis, present in OMVs from AN54 under both uninduced and antibiotic-induced conditions. While the specific proteins within each pathway differed among OMV groups, they exhibited notable similarities. Some proteins, such as malate dehydrogenase and isocitrate dehydrogenase (involved in the Krebs cycle), glyceraldehyde-3-phosphate dehydrogenase, enolase, and phosphoglycerate kinase (involved in glycolysis), were common in AN54 OMVs, both with and without induction. In contrast, in AN54Δe, many proteins were outer membrane proteins (OMPs) and proteins associated with protein synthesis, such as ribosomal subunits.

Although, the role of OMPs in antibiotic resistance in Acinetobacter baumannii is well documented, their involvement in OMVs remains incompletely understood [49]. It has been suggested that they act as a gateway for the antibiotic into the vesicular lumen, where they can be degraded by beta-lactamases such as NDM-1. Kim et al. proposed this model after observing that antibiotic-resistant E. coli OMVs had a higher abundance of OmpC and OmpF [50]. When the genes encoding the corresponding porins were deleted, the antibiotic permeability of the OMVs secreted by the mutant strains decreased. In this context, the detection of OMPs such as CarO, OmpA, OmpW, OprD, and AdeABC, together with the RND transporter, across all OMV types analyzed in our study supports the idea that these components may contribute to antibiotic influx and/or efflux processes within OMVs, potentially facilitating subsequent degradation by enzymes such as NDM-1. Interestingly, we observed a higher abundance of these OMPs in the OMVs of AN54Δe compared to the OMVs of AN54, both with and without induction. This suggests that, following the loss of the plasmid carrying blaNDM−1, AN54Δe mainly uses OMPs as a resistance mechanism, making them a major component of its OMVs protein cargo. In contrast, the lower abundance of OMPs in AN54 may be related to the energetic cost associated with the presence of the plasmid carrying blaNDM−1. The acquisition and maintenance of a plasmid imply an energetic cost that can compromise cell viability. To ensure its persistence, mutual adaptation and coevolution occur in the bacterial hosts [51]. Xiang et al. [52] demonstrated that when E. coli acquired a plasmid carrying blaNDM−5, it reduced the expression of outer membrane proteins while increasing the plasmid copy number. Therefore, it is likely that the lower abundance of OMPs in the OMVs of AN54 results from a metabolic adjustment to compensate for the energetic cost of replicating the resistance plasmid.

On the other hand, although plasmid replication may be prioritized over OMP production, we did not detect NDM-1 in OMVs from AN54 under noninduced conditions; however, it was present in OMVs obtained after induction. This may be because the NDM-1 concentration in the OMVs was below the detection limit of the MS/MS technique. Nonetheless, its presence was evidenced by carbapenemase activity in our phenotypic test. Prior induction with imipenem increased NDM-1 concentration in OMVs, allowing its detection by the MS/MS technique.

Besides, ADC β-lactamase was detected in OMVs from AN54Δe, confirming that cephalosporinases are also y transported by OMVs, even without induction. The presence of ADC, a chromosomal class C β-lactamase, in the OMVs of the control strain (which lost the resistance plasmid) and not in the OMVs of AN54 with or without induction can be explained by the prioritization of plasmid replication over other genes, such as chromosomal β-lactamases, resulting in lower ADC levels. The presence of ADC in OMVs has been previously reported by Kesavan et al. [21] in A. baumannii.

After confirming that OMVs released by AN54, both with and without induction, exhibed carbapenemase activity and that NDM-1 was present in one of them, we performed antibiotic protection assays using the OMVs on susceptible isolates. The results of this test indicate that the resistance phenotype in these strains is transient and depends on OMVs’ carbapenemase activity rather than on genetic or protein transfer. Our results are consistent with those reported by Gonzales et al. [18], who used OMVs from E. coli harboring active NDM-1 in co-culture with a susceptible of E. coli strain and observed resistance to imipenem and cefotaxime. This phenotype was not maintained after subculture. On the other hand, AN54Δe was the only strain that was successfully transformed to acquire permanent resistance via uptake of the resistance plasmid, likely due to the close similarity between the genetic backgrounds of AN54 and AN54Δe.

Given that we established conditions for primarily recovering type B OMVs, we hypothesize that packaging the entire plasmid within the vesicular lumen after carbapenem induction may be attributable to plasmid over-replication, a strategy used by AN54 to manage immediate antibiotic stress. Further studies are required to confirm this hypothesis. On the other hand, we believe this result is not due to the production of type E OMVs, since AN54’s MICs for imipenem and meropenem are > 128 µg/ml, as previously reported by Bello-López et al. [26]. Our findings align with those of Rumbo et al. and Charteejee et al., who, using OMVs from various A. baumannii isolates, successfully transferred plasmids ranging from 11 kb to a megaplasmid of ~ 122 Kb to susceptible strains [2, 16].