Introduction
Water environments are major pools of antibiotics/antibiotic resistant (AR) bacteria (Baquero et al., 2008) but the knowledge on the role of aquatic microorganisms on the dissemination/emergence of AR genes is still an under evaluated issue. Cyanobacteria are ubiquitous in aquatic ecosystems (Badger et al., 2006) and although they are exposed to antibiotic pollution/resistance (Martinez, 2009) it is not known if they may play a role on AR dissemination in natural ecosystems. Some reasons lead us to hypothesize that cyanobacteria may harbor AR genes: (1) they contain mobile genetic units such as transposable elements and plasmids (Chen et al., 2008; Christiansen et al., 2008; Lin et al., 2011), a requisite for horizontal gene transfer, the main mechanism of AR genes dissemination between distinct microbiota (Wright, 2007); (2) some cyanobacterial strains exhibit antibacterial activity (Madhumathi et al., 2011; Martins et al., 2011) and, as such, should have evolved mechanisms to protect themselves from their toxic action, i.e., they are “hot spots” for the evolution of AR (Wright, 2007); (3) some cyanobacterial strains were described as resistant to some antibiotics, including penicillin and amplicilin (Prasanna et al., 2010) and a penincilin-binding protein gene has been recently found in the cyanobacterium Thermosynechococcus elongatus (Urbach et al., 2008), though it remains to explain its weak β-lactamase activity and its physiologic role 4) it has been speculated that plasmids might determine cyanobacterial resistance to antibiotics (Chen et al., 2008), although this issue was never deeply evaluated.
The investigation of AR in bacteria usually integrates the evaluation of the susceptibility pattern to antibiotics and the search for resistance genotypes in those strains exhibiting a positive resistance phenotype. Several antimicrobial susceptibility testing approaches are standardized for the majority of bacterial pathogens (Mayrhofer et al., 2008), such as the Disk Diffusion (Matuschek et al., 2014) and Broth (Micro) dilution methods (ISO, 2006). Additionally, breakpoints or guidelines for interpreting the results from those tests are also harmonized among countries (CLSI, 2014; EUCAST, 2014) which facilitate the identification of resistant strains and the evaluation of emergence/dissemination of resistant bacterial pathogens worldwide.
Conversely, this is not the scenario for cyanobacteria since the issue of AR in cyanobacteria has been rarely investigated. The majority of studies regarding the effects of antibiotics on cyanobacteria are most related with the impact of antibiotic pollution upon aquatic ecosystems, some reporting harmful effects upon cyanobacteria (Halling-Sørensen et al., 2000; Pan et al., 2008; van der Grinten et al., 2010), but others reporting no deleterious effects (Stoichev et al., 2011). The relationship between antibiotics and cyanobacteria has also been addressed in the context of the establishment of axenic cultures. In fact, cyanobacterial cultures maintained in laboratory are usually non-axenic (Hong et al., 2010) since many water/environmental bacteria (ecosymbionts) are tightly attached to the mucilage of cyanobacteria colonies (Shiraim et al., 1989). Several methods have been proposed to obtain axenic cultures, generally involving the use of antibiotics or other chemical agents, UV radiation and physical separation approaches (Shiraim et al., 1989; Hong et al., 2010). However, the success of these purification procedures largely depends on the cyanobacteria isolates and their respective contaminants. On the other hand, testing cyanobacteria in solid media turns often a difficult task due to bacterial (over) growth since cyanobacteria exhibits lower growth rates than bacteria, within the range of 0.3–1.4 doubling per day (Mur et al., 1999). Additionally, cyanobacteria do not growth well on agar plates probably because agar might contain substances that inhibit cyanobacteria growth (Ferris and Hirsch, 1991; López-Rodas et al., 2006). Consequently, the methodologies standardized for susceptibility testing in bacteria are hardly applicable to cyanobacteria.
The information of cyanobacteria susceptibility to antibiotics is much dispersed and the available results cannot be easily compared given that very distinct experimental conditions and endpoints have been employed. In fact, cyanobacteria strains from very distinct species/habitats have been tested in distinct media, inocula, and temperatures; different antibiotic types and concentrations were used; a great variety of endpoints were employed to evaluate the effect of antibiotic on cyanobacterial growth such as chlorophyll/protein content; photosynthetic yield, cell number, optical density, among others. Besides, the results have also been expressed by different ways such as EC50, NOEC (no observed effect concentration) and resistant/susceptible phenotype (Reynaud and Franche, 1986; Liu et al., 2012; Kolar et al., 2014).
In this work we aimed to evaluate the susceptibility of several cyanobacteria species to antibiotics from distinct classes: β-lactams (amoxicillin, ceftazidime, ceftriaxone), aminoglycosides (gentamicin, kanamycin), quinolones (norfloxacin, nalidixic acid), tetracycline and trimethoprim, commonly used in the treatment of human/animal infectious diseases. We based our procedure on the conventional broth microdilution method used in antimicrobial susceptibility testing for bacteria (ISO, 2006), but optimized for the cyanobacteria culturing conditions. Data concerning the antibiotic residues in surface freshwaters is not well documented in Portugal, neither in other countries. However, the available information from environmental water samples revealed that antibiotics concentrations range from ng/L up to mg/L (Almeida et al., 2014), depending on the antibiotic and/or the sampling site (surface water, wastewater effluent, etc…). Among the above cited antibiotics we only found reported concentrations in Portugal for tetracycline (0.12–7 μg/L) (Novo et al., 2013; Varela et al., 2014), quinolones (0.03–4.4 μg/L) (Seifrtová et al., 2008; Novo et al., 2013) and trimethoprim (15.7 ng/L) (Madureira et al., 2010). On the other hand, few previous studies concerning the impact of antibiotics on cyanobacteria used the concentration range of mg/L (Vázquez-Martínez et al., 2004; Hong et al., 2010; Kolar et al., 2014). Based on these facts, we exposed the cyanobacteria to serial dilutions of the tested antibiotics within a concentration range of μg/L to mg/L, considering the worst scenario for the antibiotic contamination of freshwaters.
Materials and methods
Cyanobacterial strains
Strains of cyanobacteria, belonging to the “Estela Sousa e Silva Algae Culture Collection (ESSACC),” hereafter referred as LMECYA strains (Laboratory of Biology and Ecotoxicology, National Institute of Health, Portugal) were studied. LMECYA deposited in ESSACC have been characterized phylogenetically (Valério et al., 2009) and toxicologically (Pereira et al., 2001, 2004) and many have been used as reference strains in diverse studies (Paulino et al., 2009; Ballot et al., 2010; Martins et al., 2013).
The susceptibility to antibiotics was first evaluated in four strains from distinct genera. These strains were Microcystis aeruginosa (LMECYA 7), Aphanizomenon gracile (LMECYA 40), Chrisosporum (Anabaena) bergii (LMECYA 246), and Planktothrix agardhii (LMECYA 260), previously isolated from Portuguese freshwater resources with distinct geographical/utility characteristics: Montargil reservoir (rural area/recreational and agriculture activities), Crato reservoir (rural area/recreational and agriculture activities), Jamor lake (urban area/recreational activities), and São Domingos reservoir (rural area/agriculture and public supply), respectively. In a second step, the susceptibility to antibiotics was evaluated in eight additional strains of the same specie: the M. aeruginosa; these strains were isolated from rural freshwater reservoirs such as Montargil (LMECYA 91B, LMECYA 113, LMECYA 142), Roxo (LMECYA 50), Crato (LMECYA 108), Monte da Barca (LMECYA 151), Magos (LMECYA 159), and Corgas (LMECYA 167). All the isolates have been successfully maintained in the laboratory culture chamber as monoalgal, free of eukaryotes, non-axenic stock cultures, in Z8 medium (Skulberg and Skulberg, 1990), in a 14/10 h L/D cycle (light intensity 16 ± 4 μEm−2s−1, approx.) at 20 ± 1°C.
Antibiotics preparation
In this study we tested the following classes of antibiotics: β-lactams (amoxicillin, ceftazidime, ceftriaxone), aminoglycosides (kanamycin, gentamicine), quinolones (nalidixic acid, norfloxacin), trimethoprim and tetracycline. Concentrated antibiotic stock solutions (25x–100x) were prepared by dissolving the powder antibiotics in sterile ddH2O, according to their solubility. These stock solutions were kept at −20°C. Antibiotic work solutions were prepared immediately before used and were diluted in ddH2O to the final concentration of 6.4 mg/L.
Antibiotic susceptibility of four cyanobacteria genera
We used a methodology based on the standard microdilution method for bacteria (ISO, 2006), and adapted to the specific culturing conditions of cyanobacteria (LMECYA 7, LMECYA 40, LMECYA 246, and LMECYA 260), to evaluate their susceptibility to the antibiotics referred in Section Antibiotics Preparation. Briefly, we prepared a 96-well microplate for each antibiotic containing 100 μL of Z8 medium (Skulberg and Skulberg, 1990) in each well. Then, 100 μL of the antibiotic working solution (6.4 mg/L) was serial diluted (1/2) from column 1 to column 11 of the microplate (see Supplementary Figure 1). In parallel, 100 μL of Z8 medium were added to column 12 for the non-exposed cyanobacterial cells (control wells). Finally, 100 μL of each cyanobacterial isolate was inoculated in one microplate raw. We tested two inocula for each isolate corresponding to a final concentration of 5 × 105 cells/mL (inoculum 1) and 2 × 106 cells/mL (inoculum 2). Inoculum 1 corresponds to a biomass density within the range recommended by the OECD Guidelines for Testing of Chemicals in Freshwater Alga and Cyanobacteria (OECD, 2011). Considering that the effect of antibiotics might depend on cell densities, we also tested a four-time higher initial biomass (inoculum 2). These are cell densities commonly found in cyanobacterial blooms (Codd et al., 2005). The inocula were prepared by diluting cyanobacterial stock cultures in Z8 medium, according to their cell densities determined in Sedgewick-Rafter chambers by microscope counting (LeGresley and McDermott, 2010). Microplates were incubated in culture chamber at 20°C, 14/10 h light/dark cycle, light intensity 16 ± 4 μEm−2s−1.
Cyanobacterial cell density (OD450nm) (Churro et al., 2009) and microscopic examination of cultures integrity were followed for 13 days after incubation. Minimum inhibitory concentration (MIC) was considered as the lower antibiotic concentration that totally inhibited cyanobacteria cell growth (according to OD450nm measures in relation to control wells and corresponding to the absence of undamaged cyanobacterial cells under microscopic examination). Few cyanobacterial colonies were still observed under microscope in some wells corresponding to putative MIC, although these wells exhibited negligible OD450nm and no visible growth at naked eye. Consequently, we had to verify if those remaining cells were not viable in order to confirm the MIC value. For that purpose, 20 μL of all the microplate wells were re-inoculated in new microplates containing 180 μL of fresh Z8 medium per well and the cell growth was followed overtime (by OD450nm measurement and microscopic examination) (confirmation step 1). We considered that the MIC value was maintained if no cyanobacterial growth was observed for that antibiotic concentration. If cell growth was still observed, we considered that the MIC value corresponded to the antibiotic concentration immediately above that inhibited cyanobacterial growth. Another growth control step was performed in order to confirm all the MIC values. For that purpose, the content of the wells corresponding to the putative MICs were re-inoculated in 4 well plates (dilution 1/2 in fresh Z8 medium) and the absence of cell growth was confirmed overtime (confirmation step 2). All the procedure was repeated in two independent experiments.
Antibiotic susceptibility of eight <italic>M. aeruginosa</italic> strains
The same procedure (Section Antibiotic Susceptibility of Four Cyanobacteria Genera) was used to determine the MIC of antibiotics among the eight additional M. aeruginosa strains (referred at Section Cyanobacterial Strains), in order to evaluate antibiotic susceptibility trends within this specie. In this case, we tested the inoculum 1 in three independent experiments.
Antibiotics quality control
To ensure that the antibiotics maintained their activity during the experiment course time, we performed a stability assay using the bacterial standard strains Escherichia coli (ATCC 25922) and Staphylococcus aureus (ATCC 29213), commonly used as quality controls of Antimicrobial Susceptibility Testing (EUCAST, 2014). For that purpose, we prepare 96-well microplates containing serial dilutions of the antibiotics stock solutions in Z8 medium within the respective MIC's range of those bacterial controls (Hakanen et al., 2002; ISO, 2006). These microplates were maintained in the cyanobacterial culture chamber conditions stated at Section Cyanobacterial Strains. Samples from those antibiotics were taken at 0, 1, 7, and 14 days and used to determine the MICs of E. coli and S. aureus strains by the standard Broth Microdilution procedure and interpreted according to the EUCAST guidelines (ISO, 2006; EUCAST, 2014).
Discussion
Cyanobacteria are primary producers in freshwater ecosystems, playing important roles such as oxygen production, nitrogen fixation and nutrients supply (Badger et al., 2006; González-Pleiter et al., 2013). Considering that cyanobacteria might be particularly vulnerable to water contaminants (López-Rodas et al., 2006; González-Pleiter et al., 2013) it becomes crucial to evaluate how these organisms, and consequently the aquatic ecosystems, are impacted by antibiotic pollution. In fact, there have been an increasing concern with the ecological risk of antibiotics in the aquatic environment (van der Grinten et al., 2010) and cyanobacteria have been recommended as a test organism to predict the concentration of antibiotics in the environment for which adverse effects are not expected to occur (EMEA, 2006). This recommendation was based on the assumption that cyanobacteria are sensitive to antibiotics. However, the exposure of cyanobacteria to antibiotics has never been approached from the standpoint of resistance to antibiotics, at least in a systematic way, and the role of cyanobacteria on water resistome was never evaluated.
In this work, we observed that the response of freshwater cyanobacteria to antibiotics depends on the type/concentration of the antibiotic as well as on cyanobacteria isolate, using conditions corresponding to a realistic scenario of bloom formation and of antibiotic contamination levels in freshwater.
Our results demonstrated that the β-lactams were the most effective antibiotics against three of the four tested cyanobacteria genera, exhibiting the low MIC values (from 0.003 to 1.6 mg/L). This agrees with previous studies pointing out deleterious effects of β-lactams on cyanobacterial strains. For example, the 50% effective concentration (EC50) of amoxicillin in M. aeruginosa growth rate after 7 days of exposure was reported as 0.0037 mg/L (Lützhøft et al., 1999) and 0.008 mg/L (Liu et al., 2012). Although these values were obtained by different endpoints of toxicity (cell density and chlorophyll, respectively), they falls within the range of the MICs determined in our study. Additionally, it was described that cyanobacterial cell wall can be disrupted by the enzymatic actions of penicillin (Holm-Hansen, 1968) and that amoxicillin can impair photosynthesis in cyanobacteria (Pan et al., 2008). Conversely, it was found that the marine Phormidium valderianum is able to use ampicillin as a nitrogen source, which confer it a resistance phenotype to this antibiotic up to 2 mg/mL (Prabaharan et al., 1994). The resistance to ampicillin, carbenicillin, and penicillin (10 mg/L) was also reported for Gloeocapsa sp. and Chroococcidiopsis sp. (Reynaud and Franche, 1986). Besides, the ability to produce β-lactamases was hypothesized for Lyngbya spiralis, Anabaena variabilis and Calothrix membranacea and T. elongatus (Urbach et al., 2008; Padmapriya and Anand, 2010). A screening with Anabaena sp. and Nostoc sp. using the agar double layer method showed that both susceptible and resistant patterns were observed for some β-lactams (30 μg), depending on the strain and/or the antibiotic (Prasanna et al., 2010). The available information is still somehow contradictory and the effects of β-lactams on cyanobacteria are far for being elucidated.
According to our results, aminoglycosides exhibited MICs ranging from 0.1 to 1.6 mg/L with no apparent differences between cyanobacteria. These results agree partially with previous data. In fact, Cameron and Pakrasi (2011) showed that Synechocystis sp. is susceptible to gentamicine (1–10 mg/L). However, a resistance phenotype was shown for Synechococcus sp. exposed to 10 mg/L (Reynaud and Franche, 1986). Similarly, susceptibility to 30 μg of kanamycine was reported for Anabaena sp., Nostoc sp., Synechochoccus sp. and Pseudoanabaenacea (Lorenz and Krumbein, 1984; Prasanna et al., 2010), but studies related with the axenic purification of marine cyanobacteria have shown that the growth of Nodularia sp. (Hong et al., 2010) and Phormidium animalis (Vázquez-Martínez et al., 2004) was maintained after exposure 100 and 150 mg/L of kanamycine, respectively. It is known that, besides their principal mechanism of action (inhibition of protein synthesis), aminoglycosides can also cause cytotoxicity through the induction of reactive oxygen species (Cameron and Pakrasi, 2011). Recently, it was suggested that glutathione contributes to gentamicine resistance in the cyanobacterium Synechocystis sp. (Cameron and Pakrasi, 2011). In fact, cyanobacteria encode specific glutathione transferases and have high levels of citosolic gluthathione (Wiktelius and Stenberg, 2007). However, if glutathione may underlie the resistance of few cyanobacteria to aminoglycosides, it does not explain the susceptibility of other strains.
A review of the literature also shows that different degrees of susceptibility to tetracyclines were reported for some cyanobacteria. Anabaena spp., Nostoc sp., and Synechochoccus sp. were described as susceptible to 30–50 μg/disk of tetracycline (Lorenz and Krumbein, 1984; Prasanna et al., 2010). According to toxicity tests performed by OECD guidelines (OECD, 2011), an EC50(72h) value of 2.7 mg/L for oxitetracycline was calculated for Anabaena flos-aquae (Kolar et al., 2014) and an EC50(7days) of 0.09 mg/L for tetracycline was found in M. aeruginosa (Halling-Sørensen, 2000). On the other hand, a study showed that the photosynthetic efficiency M. aeruginosa was not affected by oxytetracycline in the range of 0.001–1 mg/L (van der Grinten et al., 2010). Some of this data fall within the MIC range for tetracycline obtained with our LMECYA isolates (some within the range of 0.1–1.6 mg/L; others with MIC > 1.6 mg/L).
Our results led us to hypothesize that the cyanobacteria might be intrinsically non-susceptible to trimethoprim and nalidixic acid (MIC > 1.6 mg/L). Trimethoprim exerts its antibiotic activity by inhibiting the dihydrofolate reductase (folA), an essential enzyme involved in folate metabolism in prokaryotes (Myllykallio et al., 2003). However, it was previously demonstrated that several bacteria and two strains of cyanobacteria (Nostoc sp. PCC7120 and Synechocystis sp. PCC6803) lacks genes encoding folA and that they have an alternative pathway (thymidylate synthase, ThyX) to synthesize reduced folate molecules required for intermediary metabolism (Myllykallio et al., 2003). Inclusive, strains of Helicobacter spp. and Campylobacter spp. lacking folA and using ThyX for nucleotide synthesis were considered endogenously resistant to low levels of trimethoprim (Myllykallio et al., 2003). If the presence of alternative pathways to folA is a common feature of cyanobacteria, then we can postulate that cyanobacteria are naturally non-susceptible to trimethoprim. Actually, previous studies have shown that some cyanobacteria exhibit reduced susceptibility to this antibiotic. A microplate test with M. aeruginosa revealed that cyanobacterial growth is not inhibited after 24 h of exposure to trimethoprim within the range of 0.001–10 mg/L (EC50 = 6.9 mg/L) using the photosynthetic yield as endpoint of antibiotic effect (van der Grinten et al., 2010). A considerable higher EC50 value for trimethoprim (after 72 h of exposure) was also reported for M. aeruginosa (112 mg/L) (Halling-Sørensen, 2000) as well as for Anabaena flos-aque (253 mg/L) (Kolar et al., 2014) [obtained by toxicity assays according to OECD guidelines (OECD, 2011) for algal growth inhibition test]. Another study demonstrated that an Anabaena sp. and two Nostoc spp. strains were resistant to 25 μg of trimethoprim after exposure to antibiotic disks for 2–4 days on 0.8% agar medium (Prasanna et al., 2010). These studies can be hardly compared considering the differences in the experimental procedures, but they point out the reduced susceptibility of cyanobacteria to trimethoprim.
In our study we found that the four tested cyanobacteria genera were not susceptible to nalidixic acid (within the range of 0.0015–1.6 mg/L), but they exhibited distinct susceptibility degrees to norfloxacin, being M. aeruginosa particularly susceptible to this antibiotic (0.012 mg/L ≤ MIC ≤ 0.2 mg/L). The distinct response to these two antibiotics is not surprising considering that, among bacteria, nalidixic acid is a less potent quinolone than norfloxacine (Andersson and MacGowan, 2003). The bactericidal activity of quinolones occurs by the inhibition of the DNA gyrase and DNA topoisomerase IV, essential enzymes involved in DNA replication, and the mechanisms of quinolones resistance are due to modifications of these target enzymes or to changes in the antibiotic entry/efflux (Andersson and MacGowan, 2003; Jacoby, 2005). In Gram-negative bacteria the reduction of membrane permeability can be achieved by changes in the expression of porins (Jacoby, 2005). The cell wall from cyanobacteria has an overall structure similar to the Gram-negative cell wall, but their peptidoglycan layer is much thicker and their porins has a lower conductance (Hoiczyk and Hansel, 2000). The low cut-off of cyanobacterial porins is related with the protection against harmful agents such as toxins and antibiotics (Hoiczyk and Hansel, 2000). Considering these facts, we might hypothesize that the apparent reduced susceptibility to nalidixic acid of the four cyanobacteria genera we studied is due to their low permeability to this antibiotic. Besides, we also might consider that differences in cell wall permeability might also explain the higher susceptibility of M. aeruginosa (LMECYA 7) to norfloxacin, in comparison with other cyanobacteria, particularly the P. agardhii (LMECYA 260). It is known that peptidoglycan layer of cyanobacterial cell walls varies among genus/species, being considerable thicker in species from the Oscillatoriales order, to which P. agardhii belongs. Two previous studies have also reported that a strain of Anabaena sp. (Lorenz and Krumbein, 1984; Prasanna et al., 2010) and a strain of Synechochoccus were resistant to 30 μg/disk of nalidixic acid. However, the same authors also described a susceptible phenotype to the same dose of this antibiotic for another strain of Anabaena sp. and two strains of Nostoc sp. (Prasanna et al., 2010), as well as for a strain of Pseudoanabaena sp. (Lorenz and Krumbein, 1984). As for the other antibiotics, a definitive conclusion concerning the effects of quinololes on cyanobacteria is not possible to achieve with the available data.
Considering the cyanobacteria strains we conclude that M. aeruginosa (LMECYA 7) was the less susceptible to amoxicillin, but the most susceptible to norfloxacin. A. gracile (LMECYA 40) and C. bergii (LMECYA 246) were particularly susceptible to β-lactams. P. agardhii (LMECYA 260) were most susceptible to amoxicillin, but exhibited moderate to high MICs for other antibiotics. It is tempting to associate these differences with the cyanobacterial order since M. aeruginosa belongs to Chroococcales, A. gracile and C. bergii belong to Nostocales, and P. agardhii to Oscilatorialles orders. In fact, the consistency observed with the results from the nine tested M. aeruginosa strains (that were isolated from distinct freshwater reservoirs), supports the hypothesis that cyanobacteria specie might be related with specific antibiotic susceptibility patterns. On the other hand, the observation that all tested cyanobacteria were not susceptible to trimethoprim and nalidixic acid (from 0.0015 to 1.6 mg/L) supports the hypothesis that cyanobacteria, independently of the specie, might share common responses to antibiotics. However, further studies using a higher number of strains for the other species will be required in order to identify putative antibiotic susceptibility patterns related with cyanobacteria specie, genus or orders.
In conclusion, the reduced susceptibility of cyanobacteria to some antibiotics suggests that they may be naturally (constitutively) non-susceptible to these compounds or even that they might acquire antibiotic non-susceptibility due to environmental selection pressure by antibiotic exposure or to the transference of antibiotic resistance genes from bacteria. Considering that the available information on this issue is very dispersed and difficult to compare studies including species from a diversity of water bodies will be fundamental to map the resistance phenotypes/genotypes of cyanobacterial isolates from freshwater environments. Another fundamental aspect will be the establishment of standard procedures and breakpoints in order to define the MIC values to the different classes of antibiotics for cyanobacteria.
The knowledge on the role of cyanobacteria on the water resistome will help to understand how the aquatic ecosystems react to antibiotic pollution and to define preventive/remedial measures concerning the dissemination of antibiotic resistance in the environment.