Zebrafish embryo is an appropriate model to monitor a persistent infection with P. aeruginosa CF isolates

While reference laboratory strains causing acute infections (PAO1, PA14, and PAK) have been widely used in the zebrafish model to gain insights into the host-P. aeruginosa interaction in vivo, investigations of the pathogenesis of clinical isolates in this vertebrate model have remained very scarce, and relied on the microinjection of single CF isolates [21,22]. We recently developed a wound infection protocol in zebrafish based on the immersion of tail fin-amputated embryos with P. aeruginosa PAO1 strain, which caused an acute infection within 20 hours post infection (hpi) [23]. The immersion of injured embryos is a mode of infection that is easier than microinjection, reproducible, amenable to confocal and electron microscopy and reflecting a natural infection by P. aeruginosa. Here, we used this infection mode to evaluate the in vivo virulence of three CF isolates (A6520, B6513 and C6490), that were not tested earlier for virulence or persistence in any infection model. Genome sequencing indicated that these three strains, all positive for exoS gene, belong to distinct serotypes (ST274, ST27 and ST633). The survival of zebrafish embryos was first monitored over 40 hpi (Fig 1A). All three CF isolates were highly attenuated, being associated with 90% (A6520, B6513) to 100% (C6490) survival, when compared to the reference strain PAO1 (50% survival) (Fig 1B). Additional experiments were next conducted on these CF isolates to assess whether their attenuated phenotype was linked to the elimination or not of the bacteria by the host.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. CF isolates can establish a persistent infection in zebrafish embryos, with parallel profiles to a murine model of infection.

(A) Experimental timeline to assess bacterial virulence and persistence in zebrafish embryo. Drawing was created with BioRender.com. (B) and (D) Embryo survival following infection by immersion with indicated P. aeruginosa strains at bacterial concentrations ranging from 7.2 x 10⁷ to 1.2 x 10⁸ CFU/mL. Survival was monitored for > 40h following the infection (n = 3, 60 larvae in total). Log-rank test: *P < 0.05, **P < 0.01 and ****P < 0.0001. (C) Evolution of the bacterial load per embryo over time (until 72 hpi). Following infection by immersion with GFP⁺ P. aeruginosa isolates, embryos were crushed at the indicated time points and plated for CFU counting (n = 3, 15 larvae). (E) and (F) Virulence and persistence of CF isolates in a murine model of high-density cutaneous infection. Bacterial load (E) and size of abscesses (F) formed for three or seven days in CD-1 mice subcutaneously injected with indicated P. aeruginosa strains in the right dorsum were quantified. The CFUs reported at 0 dpi (empty squares) correspond to the injected inoculum. At 3 dpi and 7 dpi, abscesses were measured and harvested in phosphate buffered saline (PBS), homogenized and plated on lysogeny broth (LB) for bacterial enumeration. Data from two independent experiments containing 2–4 biological replicates each (n = 5–7) are displayed as the median. The limit of detection (LOD) is displayed as a dashed line at 10² CFU/abscess. The photo insets above the graph are representative images from treatment groups (M.A.A is the photographer).

https://doi.org/10.1371/journal.ppat.1012922.g001

To discriminate between bacterial clearance and bacterial persistence, we assessed the evolution of the bacterial load per embryo over 72 hpi. To easily recognize and count P. aeruginosa bacteria, we used green-fluorescent strains harboring a chromosomally-integrated gfp gene with constitutive expression, to avoid the risk of GFP signal loss upon time. For all strains, a notable variability between individual larvae was noticed, with some differences being > 100-fold (2 log₁₀), but the distribution profiles were reproducible between independent experiments. In surviving embryos infected with the reference strain PAO1, a continuous decrease of bacterial load was found over time but some bacteria were persisting (Fig 1C), which is consistent with a previous study using a different infection mode [24]. With isolates A6520 and C6490, the majority of the embryos were observed to be free of bacteria at 67 or 43 hpi, respectively (Fig 1C). Conversely, with strain B6513, 90% of the bacteria were eliminated between 1.5 to 18 hpi, but the remaining bacterial load (median of 630 CFU) stayed relatively constant in numbers until 65 hpi (Fig 1C). Thus, isolate B6513 developed a persistent infection in this immersion zebrafish model, whereby a subset of bacteria survived after 18 hpi without killing the host. The residual CFU analysis for isolate B6513 was extended up to 6 days post-infection (dpi) and an increased number of persistent bacteria per fish was observed compared to 3 dpi. However, further studies would be required at 6 dpi to differentiate the persistent bacteria from possible reinfections through mouth ingestion. Subsequent experiments were carried out only on embryos for up to 3 dpi (i.e., 5 days post-fertilization, within the frame not regulated as animal experiments) to follow the 3R-principle and limit a possible reinfection upon food ingestion.

To extend the relevance of this model, we assessed the virulence and persistence of the well characterized isolate RP73, a late CF isolate capable of long-term colonization in mouse airways [25,26]. Consistent with previous mouse studies, and similarly to strain B6513, the RP73 isolate failed to induce embryo mortality and persisted in infected zebrafish embryos (Fig 1C and 1D). To verify that the persistent phenotype reflected colonization of the wound and not the ability of the isolates to adhere to the embryo’s skin, we assessed the bacterial load over time in uninjured embryos incubated with persistent strains B6513 and RP73. At 24 hpi, no CFU were detected for either isolate (S1 Fig), showing that the injury was required to drive a persistent infection.

We next addressed the behaviour of the three CF isolates A6520, B6513 and C6490 in a mouse model of chronic infection that involves cutaneous abscesses [16]. P. aeruginosa LESB58, a well-characterized CF isolate that causes chronic lung infection and chronic skin abscess in mice [16,27], was used as a reference to assess in vivo growth and virulence of the clinical isolates over a seven-day period after subcutaneous injection in mice (with approx. 5 x 10⁷ CFU). As previously reported, P. aeruginosa LESB58 persisted at the infection site, being recovered at median densities above 10⁷ CFU/abscess at 3 and 7 dpi (Fig 1E). Similarly, the clinical isolate P. aeruginosa B6513 was recovered at high median densities, > 10⁸ CFU/abscess at 3 and 7 dpi. In contrast, clinical isolates P. aeruginosa A6520 and C6490 were mostly eliminated from the host when compared to the other strains, as reflected by low median bacterial densities (< 10⁴ CFU/abscess at 3 dpi, and 10³ or below the limit of detection at 7 dpi). Trends in abscess sizes were reflective of bacterial recovery from the localized infection site at both time points, with abscess (dermonecrotic cutaneous tissue) lesion areas of 15–39 mm² and 47–98 mm² for strains LESB58 and B6513, respectively, whereas strains A6520 and C6490 did not cause abscess formation (Fig 1F). Healing of the cutaneous tissue appeared to occur at 7 dpi for both P. aeruginosa LESB58 and B6513, since the sizes of abscesses had decreased (−24 and −51 mm² respectively) relative to the abscesses of animals sacrificed 3 dpi. Accordingly, the mean clinical scores of animals decreased during the time from 3 to 7 dpi across treatment groups (S2 Fig). Furthermore, clinical scores of animals remained low overall, with a maximum recorded total score of 8/50 (in the P. aeruginosa B6513 treatment group), indicating that none of the strains caused severe disease in animals.

Taken together, our results showed that the injured embryo immersion model of zebrafish infection is suitable to assess persistence of P. aeruginosa clinical isolates. Importantly, the persistence profiles of strains in zebrafish are consistent with the phenotypes observed in a murine chronic infection model. We next took advantage of the unique imaging opportunities offered by the zebrafish embryo to visualize persisting bacteria at high resolution.

Dynamic interaction of persistent P. aeruginosa with host immune cells in vivo

Macrophages and neutrophils have been shown to rapidly phagocytose P. aeruginosa upon zebrafish infection [20,28–30]. We took advantage of the optical transparency of embryos to image GFP-expressing isolate B6513 throughout the infection, assess bacterial morphology, localization and interaction with recruited macrophages, using Tg(mfap4::mCherry-F) zebrafish embryos that harbor red fluorescent macrophages. Consistent with CFU counts, we observed a decrease in the GFP⁺ bacteria between 1.5 and 24 hpi (Fig 2A). At 1.5 hpi, most bacteria appeared outside of mCherry labeled macrophages, while at 24 and 65 hpi, they mainly appeared as foci localized nearby the injury site (Fig 2A). Intra-macrophagic P. aeruginosa were present throughout the course of infection (Fig 2A) and a 3D reconstruction is shown in Fig 2B (the sequential Z stack slices of the 3D reconstruction are shown in S3 Fig). The presence of bacterial foci, with putative localization inside macrophages, was not specific to the B6513 isolate since similar findings were visualized with the RP73 strain (S4 Fig). The rare persistent bacteria visualized with PAO1 strain can also be found inside macrophages (S4 Fig), as reported earlier using a different infection mode [24]. Bacterial aggregates were retained within macrophages for 10 h or more, as shown using time-lapse confocal microscopy (Fig 2C). The size distribution of GFP positive spots was stable over time, with a volume mainly comprised between 1.5 µm³ (individual bacteria) and 100 µm³ (clusters) in size (Fig 2D). Larger foci (> 100 µm³) were observed in some embryos. Infection seemed to delay macrophage recruitment to the wound at the early time point (1.5 hpi, Fig 2E). Macrophage recruitment in infected embryos peaked at 24 hpi and decreased slowly, being close to the level of uninfected controls at 72 hpi (Figs 2E and S5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Upon persistent colonization, P. aeruginosa B6513 forms aggregates which can be visualized inside macrophages but not neutrophils.

(A) Representative maximal projections of confocal images, showing interactions between bacteria (green) and recruited macrophages (red) in Tg(mfap4:mCherry-F) larvae at different time points. White rectangles denote images extracted from a single optical section of macrophages with intracellular P. aeruginosa. Below the images, orthogonal representations of the (*) events are shown, confirming that bacteria were inside macrophages (only one example of orthogonal view is shown but intracellular location applies to all rectangles). Scale bar: 40 µm. Note that pictures come from different embryos imaged for each indicated times (10 to 17 larvae at each time point). (B) 3D reconstruction allowing to confirm the intra-macrophagic localization of a bacterial foci. Scale bar: 7 µm. (C) Zoomed single optical sections of an intra-macrophagic bacterial aggregate tracked from 24 hpi to 34 hpi using time-lapse confocal imaging. Image acquisition was done every 30 min and images corresponding to 2 h intervals are shown. Brightness/contrast settings were modified comparatively to (A) for a better visualization of the aggregate organization after a deconvolution. Scale bar: 5 µm. (D) Volume repartition of total GFP⁺ events per larvae quantified following 3D reconstruction (10 to 13 embryos were imaged at each time point), along with representative images (single Z-plans) of the four types of size clusters for GFP⁺ events. (E) Macrophage quantification at the wound, measured by the number of red pixels at the median plan of the stack, in presence or absence of bacteria at various time points (4 to 8 control and 8 to 18 infected embryos were imaged at each time point). Mann-Whitney test: *P < 0.05, **P < 0.01 and ***P < 0.001. (F) Representative maximal projections of confocal images, showing interactions between bacteria (green) and recruited neutrophils (red) in Tg(LysC:dsRed) larvae. White rectangles denote images extracted from a single optical section of neutrophils with intracellular P. aeruginosa. Scale bar: 40 µm. Note that pictures come from different embryos imaged for each indicated times (12 to 16 larvae at each time point). (G) Evolution of the proportion of Tg(mfap4:mCherry-F) and Tg(LysC:dsRed) larvae with at least one event of intra-macrophage or intra-neutrophil bacteria, respectively, over time (10 to 16 larvae were imaged at each time point). Only colonized embryos (presence of GFP⁺ events) with detectable macrophages or neutrophils (mCherry⁺ signal) were included in the analysis. Drawings in panels A and F were created with BioRender.com.

https://doi.org/10.1371/journal.ppat.1012922.g002

The interaction of persistent bacteria with neutrophils using the Tg(LysC:dsRed) transgenic line harboring red fluorescent neutrophils was subsequently investigated (S6 Fig). In agreement with previous experiments done on microinjected embryos [28–30], numerous intra-neutrophilic P. aeruginosa were observed shortly after colonization (Fig 2F and 2G). However, this frequency was strongly reduced at 24 and 48 hpi (S6 Fig), and no intra-neutrophil bacteria were observed at 72 hpi. On the other hand, the association between P. aeruginosa and macrophages was maintained over time with around 60% of infected embryos having at least one event of intra-macrophage bacteria (Fig 2G), whereas the remaining infected embryos displayed no intra-macrophage bacteria (S4 Fig). In embryos with at least one intra-macrophage event, the percentage of GFP⁺ foci inside macrophages was around 14%, 32% and 34% at 24, 48 and 72 hpi, respectively (high bacterial load at 1.5 h precluded image quantification of the GFP signal, but GFP⁺ foci inside macrophages were estimated below 3% at this early time point). Thus, the kinetics of appearance of intra-neutrophil and intra-macrophage bacteria were very different (Fig 2G), suggesting that the interaction with neutrophils was restricted to the early steps of colonization and/or that the killing activity of neutrophils and/or their short half-life prevented any intra-neutrophilic persistence of P. aeruginosa. Cumulatively, this live imaging analysis revealed transition from bacteria mostly outside macrophages and neutrophils to persistent clusters in an intra-macrophage niche.

High resolution imaging of the intracellular niche of persistent bacteria

As described above, live imaging revealed the presence of persistent P. aeruginosa bacteria, both inside and outside macrophages at 24 and 48 hpi (Fig 2A and 2G), but likely not within neutrophils (Fig 2F and 2G). Since P. aeruginosa can also enter and reside within non-phagocytic cells [31–33], we investigated their presence within non-phagocytic cells at the infection site. To this aim, we used Tg(rcn3:Gal4/UAS:mCherry) transgenic embryo harboring red fluorescent mesenchymal cells. Intracellular bacteria (individual or clusters) were clearly visualized at 24 and 48 hpi (S7 Fig), demonstrating that P. aeruginosa also uses this niche in vivo, supporting pioneer work carried out with mice corneal infections [9,33].

We used electron microscopy (EM) to highlight the ultrastructural context of persistent bacteria in the tail fin of 24 or 48 hpi fixed infected Tg(mfap4::mCherry-F) embryos (Fig 3). Bacteria were largely seen inside cells, either single (Fig 3A) or in clusters (Fig 3B). A detailed analysis indicated that some bacteria were rather localized in vacuoles, as shown by the shape of clusters and/or the visualization of a vacuolar membrane, whereas other appeared free in the cytoplasm (Fig 3A). Some cells with intracellular bacterial cluster appeared highly damaged (Fig 3C, left panel) and bacteria could be seen in cell remnants, containing a nucleus but essentially devoid of cytoplasm, suggesting necrotic cells (Fig 3C, middle panel). Rare bacteria could be visualized in the extracellular space (Fig 3C, right panel). Since identifying the host-cell type only by ultrastructural criteria was challenging, we also performed correlative Light-Electron Microscopy (CLEM) (Fig 3D and 3E). A single large cluster of bacteria inside a macrophage was found by light microscopy (Fig 3D). The same embryo was processed for EM and serial sections were prepared and imaged by array tomography (Fig 3E). The same cluster, highlighted by a white square in panels D and E, showed a high number of bacteria inside a cell identified as a macrophage by the corresponding fluorescence image.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Electron micrographs of persistent P. aeruginosa B6513 in zebrafish.

Representative images were acquired from serial thin sections of infected embryos at 24 hpi (D-E) or 48 hpi (A-C). (A) High magnification images of single intracellular bacteria near the tail fin edge (pixel size 5 nm). Right panels show enlargements of the white squares. Nu: cell nucleus. A membrane is visualized around some bacteria (white arrow). (B) Overview of the tail fin region (pixel size 20 nm) showing clustered intracellular bacteria in a vacuolar-like shape (black arrows). Right panels show higher magnification (pixel size 5 nm) of the white squares. (C) High magnification images of intracellular bacteria in highly damaged cell (left panel) or in necrotic cell ghost (middle panel), or of bacteria in the extracellular space (right panel). (D) and (E) Correlative Light-electron Microscopy (CLEM). (D) A z-stack of a live zebrafish embryo was acquired. The left panel shows a single fluorescent plane overlaid with the central DIC image of the fish tail (macrophages are seen in red and bacteria in green). A single large cluster containing dozens bacteria inside a macrophage is shown (white square). The right panel shows the orthogonal projection of the portion of the stack containing the cluster. (E) After processing the same fish for EM, serial thin sections were prepared and imaged by EM. One section was acquired at a pixel size of 25 nm. Left panel shows an overview of the tail region. The large cluster observed in panel D in light microscopy is shown in its physiological context in the left panel (white square). The portion of the image containing the cluster is enlarged on the right panel, showing a high number of bacteria, filling almost all the cytoplasm of the cell and leaving the nucleus intact on one side.

https://doi.org/10.1371/journal.ppat.1012922.g003

Taken together, live and electron microscopy analyses indicated that P. aeruginosa bacteria persisting at the infection site were largely intracellular, in either macrophages or non-phagocytic cells. Moreover, CLEM imaging allowed visualization of a large cluster of P. aeruginosa inside a macrophage. These findings support the uniqueness of the zebrafish model in deciphering how P. aeruginosa interacts with host cells during infection.

Strains with a persistent profile in zebrafish survived inside cultured macrophages and formed biofilms in vitro

In our in vivo model, persistent P. aeruginosa were visualized as intra-macrophage bacterial clusters at the zebrafish wound site (Figs 2 and 3). We addressed the in vitro features of the CF isolates by investigating their phenotypes in a macrophage cell line and in a biofilm assay. We first monitored the intra-macrophagic survival of the different clinical strains by performing a gentamicin protection assay in murine J774 cells infected at a multiplicity of infection (MOI) of 10, as done previously with PAO1 [34,35]. After 1.5 h of internalization, quantification of intracellular bacteria revealed that strain A6520 was largely eliminated, while the number of C6490 and RP73 cells were only modestly reduced (Fig 4A). On the other hand, B6513 was not eliminated and survived inside macrophages for at least 1.5 h after phagocytosis (Fig 4A). Phagocytic rates of the different strains were not significantly different (Fig 4B). Time-lapse microscopy performed during 3 h after phagocytosis correlated with CFU measurements, with a striking clearance of strain A6520, not due to a disappearance of macrophages, while B6513 was still residing within macrophages (Fig 4C). The visualization of intra-vacuolar clusters with B6513 showed that this bacterium can reside in closed compartments within phagocytic cells (Fig 4D). We also assessed the cytotoxicity driven by intracellular P. aeruginosa strains 1.5 h after phagocytosis using Trypan blue [34]. CF isolates exhibited low or no cytotoxicity (6% dying cells for B6513 and below 1% for RP73, A6520 and C6490, versus 12% for the reference strain PAO1). We next evaluated the ability of the diverse strains to form biofilms. While A6520, B6513 and RP73 produced a pellicle covering the broth medium, C6490 failed to form such a biofilm at the air-liquid interface (Fig 4E). Moreover, we compared the swarming motility and pyocyanin production of A6520, B6513 and C6490 CF strains in comparison with PAO1 (non-CF strain) and LESB58 (CF strain). A6520, B6513 and C6490 strains exhibited no or strongly reduced swarming motility and low to mild pyocyanin production (S8 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Persistent isolates are able to survive inside macrophages and form biofilm in vitro.

(A) Bacterial survival within J774 macrophages (n = 4 to 8). The number of intramacrophagic bacteria was determined 20 min (T0) or 2 h after gentamycin treatment (T1). Survival is expressed as the ratio between the numbers of bacteria recovered at T1 versus T0. One-way ANOVA: ****P < 0.0001; Tukey post-hoc test: *P < 0.05, **P<0.01 and ****P < 0.0001. (B) Phagocytosis efficiency of each strain (n = 4 to 8). The number of intracellular bacteria recovered at T0 was compared to the initial inoculum. (C) Representative initial and final images of time-lapse microscopy to follow the behavior of GFP⁺ intramacrophagic bacteria (isolates A6520 and B6513) for 2h30. Scale bar: 10 µm. (D) Time lapse imaging of an intra-vacuolar bacterial cluster (red arrowhead) visualized within a macrophage (a zoom is shown below). In (C) and (D), times after the start of gentamycin treatment are indicated. (E) Biofilm formation at 24h assessed by crystal violet (CV) assay on glass tubes (n = 3 to 10). Quantification of the CV-labelled biofilm rings (a representative picture is shown on the right) is normalized with bacterial growth (OD_600nm). Kruskal-Wallis test: ***P < 0.001; Dunn post-hoc test: *P < 0.05, **P < 0.01 and ***P < 0.001.

https://doi.org/10.1371/journal.ppat.1012922.g004

Overall, these data show that the two isolates recognized as persistent in the zebrafish embryo model (B6513 and RP73) combined the ability to resist killing by macrophages upon phagocytosis and the ability to form biofilms, whereas the two non-persistent isolates lacked one of these two traits. In addition, it shows that the B6513 strain, isolated from a transient infection of a one year old child, shares several features of CF-adapted isolates (low virulence, ability to form biofilm, moderate pyocyanin production and strongly reduced swarming motility). Moreover, the serotype of B6513 (ST27) has been reported at high levels in CF patients [36]. In further studies, it will be of interest to determine how the B6513 genome and transcriptome are related to those of CF-adapted strains.

Persistent P. aeruginosa bacteria exhibit adaptive resistance to antibiotic treatment

Antibiotic adaptive resistance is a major issue in the treatment of chronic infections. We investigated antibiotic efficacy on embryos infected by the isolate B6513, by applying a 30 min treatment at 1.5, 24 and 48 hpi (Fig 5A). Three clinically-used antibiotics to fight P. aeruginosa infections were tested, namely tobramycin (an aminoglycoside), colistin (a polymyxin) and ciprofloxacin (a fluoroquinolone) [37]. The minimum inhibitory concentration (MIC) of tobramycin, colistin and ciprofloxacin for B6513 was determined experimentally to be 1 µg/mL, 4 µg/mL and 0.5 µg/mL, respectively. Antibiotics were used at 40 times the MIC (40 µg/mL for tobramycin and 20 µg/mL for ciprofloxacin), except for colistin that was toxic for embryos at 40 x MIC, but 2.5 x MIC (10 µg/mL) was tolerated by the embryos. All drugs caused a sharp reduction in bacterial loads when applied soon after the initiation of infection (1.5 hpi), from approx. 20-fold (ciprofloxacin and tobramycin) to complete eradication (colistin) (Fig 5B). However, their efficacy was strongly reduced when used at 24 or 48 hpi, with colistin and ciprofloxacin decreasing CFU counts by less than 7-fold and tobramycin having no significant effect (Fig 5B). The ability of bacteria to withstand antibiotic challenges occurred at around the time of establishment of a persistent infection in the embryos, associated with intracellular and aggregated bacteria. This phenotype was not linked to the acquisition of mutations conferring resistance, since colonies isolated from embryos following ciprofloxacin or tobramycin challenge at 48 hpi remained as sensitive as the original B6513 to these antibiotics in vitro (see procedure in Methods section). Moreover, the adaptive resistance phenotype was not specific to isolate B6513 since a similar phenotype was found with RP73 strain upon treatment with colistin (S9 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Antibiotics have a reduced efficacy on persistent P. aeruginosa in infected zebrafish.

(A) Experimental procedure used to assess the efficacy of antibiotic treatments on infected embryos. ATB 30’: antibiotic treatment for 30 min. (B) Efficacy of various antibiotics on strain B6513 with respect to the time post-infection (n = 3 to 4, 15 to 21 larvae). Embryos colonized for 1.5, 24 or 48 h were subjected to the indicated antibiotic challenge, or incubated in water for the control condition. Following this 30 min treatment, bacterial load per embryo was determined in both groups. Mann-Whitney test: **P < 0.01, ***P < 0.001 and ****P < 0.0001. Ratios were calculated regarding the median of the data set; thus there is no values for colistin at t1.5 h as the median of the treated group is zero. (C) Efficacy of ofloxacin, an antibiotic known to enter eukaryotic cells with high efficiency, on strain B6513 (same analysis as in (B)). (D) Representative maximal projections of confocal images of infected Tg(mfap4:mCherry-F) larvae (all mCherry⁺), treated or not with ofloxacin at 24 hpi. For both conditions, the number of embryos with or without at least one event of intra-macrophage bacteria are indicated below the images. Among the ofloxacin-treated embryos, 2 were GFP-negative, indicative of bacterial elimination, whereas all non-treated embryos were GFP-positive. Drawings in panels A and D were created with BioRender.com.

https://doi.org/10.1371/journal.ppat.1012922.g005

During infection, P. aeruginosa may benefit from intracellular niches to escape antibiotics [8,11]. Our microscopy analysis revealed the intracellular localization of persistent P. aeruginosa, which could de facto be protected from antibiotics (Figs 2A–C, 3 and S7). Antipseudomonal antibiotics differ in their ability to penetrate into eukaryotic cells, and their activity vary against intracellular P. aeruginosa in cultured phagocytes [11]. To assess if the intracellular niche was responsible for the treatment failure in our model, we tested the in vivo activity of another fluoroquinolone, ofloxacin, that is known to be host cell permeable and to demonstrate better diffusion into human tissues than ciprofloxacin [38]. Persistent bacteria at 24 and 48 hpi exhibited some tolerance towards ofloxacin (used at 40 x MIC, 20 µg/mL), when compared to bacteria treated very early after the start of infection. However, among the four antibiotics tested, ofloxacin was the most efficient molecule at the persistent stages (24 and 48 hpi), thus reinforcing the notion that eradication of intracellular persistent bacterial reservoirs requires administration of antibiotics able to reach these niches (Fig 5C). To complete this finding, we imaged infected Tg(mfap4::mCherry-F) embryos at 24 hpi treated or not with ofloxacin (Fig 5D). The percentage of larvae with intra-macrophage bacteria dropped below 30% in ofloxacin-treated fishes, whereas it remained around 70% in non-treated controls (as seen in Fig 2G).

Taken together, our results showed that the persistence of P. aeruginosa in zebrafish embryos is associated with an adaptive phenotype of resistance to antibiotics. Importantly, the most cell-permeant antibiotic was also the most effective one, suggesting that intracellular bacteria are reservoirs resistant to antibiotics that do not enter host cells.

Anti-biofilm compounds potentiate antibiotic efficacy on persistent bacteria

In our zebrafish model, persistent P. aeruginosa formed aggregates, which may display a higher tolerance to antibiotics than free-living bacteria [7]. We hypothesized that anti-biofilm compounds would help to disassemble bacterial aggregates, thereby re-sensitizing persistent P. aeruginosa to antibiotics, potentially promoting their eradication with the assistance of the immune system. Among commercially available molecules with appealing activity against P. aeruginosa biofilm, we first used the human Atrial Natriuretic Peptide (hANP), an efficient biofilm-disperser in vitro that potentiates different antibiotics including tobramycin [39]. Though devoid of antibacterial activity by itself, peptide hANP was able to reduce the formation of biofilms by strain B6513 in vitro (S10 Fig). We then assessed its capacity to potentiate tobramycin and colistin treatments in infected embryos at 48 hpi, a time point at which bacteria were partly tolerant to antibiotics (Fig 6A). When applied alone, hANP modestly reduced bacterial loads by 2.5-fold, suggesting that it somehow assisted the innate immune system to eliminate bacteria. Strikingly, when used in combination with antibiotics, a more significant reduction of the pathogen burden was recorded, i.e., 7-fold for tobramycin and 16-fold for colistin, when compared with the controls (Fig 6B). We performed similar experiments with a structurally different anti-biofilm compound, the fatty acid cis-2-Decenoic acid (CDA). CDA, a molecule produced by P. aeruginosa, promotes biofilm dispersal and potentiates clearance of pre-established biofilms in combination with antibiotics [40,41]. Moreover, CDA was proposed as a suitable molecule to target intra-macrophagic biofilms of pathogenic Escherichia coli [42]. When used alone at 48 hpi, CDA had no detectable effects on persistent P. aeruginosa in zebrafish embryos, but when added in combination with colistin, bacterial loads per embryo decreased by 5-fold, thus increasing the antibiotic efficacy by > 3-fold (Fig 6C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Treatment with anti-biofilm compounds can re-sensitize persistent P. aeruginosa bacteria to antibiotics in infected zebrafish.

(A) Experimental procedure to assess the combine effect of anti-biofilm molecules and antibiotics. ATB 30’: antibiotic treatment for 30 min. Drawing was created with BioRender.com. (B) and (C) Antibiotic potentiation by 10 µM hANP (B) and 20 µM CDA (C) at 48 hpi (n = 2 or 3, 9 to 15 larvae). Anti-biofilm compounds were added for 5 h to embryos infected by B6513-gfp strain, followed by the indicated antibiotic treatment, before CFU counting per embryo. One-way ANOVA: ****p<0.0001; Dunnett post-hoc test (comparison with the control condition): *P < 0.05, **P < 0.01 and ****P < 0.0001. Ratios were calculated regarding the median of the data set. (D) Maximal projections of confocal images, showing the effect of hANP or CDA following the treatment used in (B) and (C), without antibiotic treatment. Pictures of Tg(mfap4:mCherry-F) larvae infected by strain B6513-gfp were taken, from the beginning of the treatment, every 30 min for 5.5 h. A selection of images at 3 time points is shown. Scale bar: 40 µm.

https://doi.org/10.1371/journal.ppat.1012922.g006

We also tested the anti-biofilm compounds (hANP or CDA) in combination with the most potent antibiotic, ofloxacin. However, the anti-biofilm compounds did not improve the ofloxacin efficacy, which may be due to the fact that dispersed bacteria may be poorly susceptible to ofloxacin due to metabolic issues. Alternatively, it may be due to the fact that most bacterial aggregates are intracellular, possibly inside vacuoles, and may reflect an insufficient penetration of anti-biofilm molecules.

We next took advantage of the optical transparency of embryos to visualize in vivo the effect of these anti-biofilm molecules on persistent bacteria. After a 5 h exposure to hANP or CDA, a reduction of the size of some bacterial clusters and concomitant increased number of isolated bacteria were observed, which was not the case in the H₂O control conditions (Figs 6D and S11). In the few treated embryos that were imaged, we did not observe the dispersion of intra-macrophage bacterial clusters. The ability of hANP and CDA to enter cells and cross membrane vacuole in our working conditions remains unknown. Currently, we cannot determine if anti-biofilm molecules act on intracellular and/or extracellular bacteria because Tg(mfap4::mCherry-F) labeling do not allow to distinguish extracellular bacteria from intracellular bacteria residing in non-macrophage cells. One can also hypothesize that bacterial clusters from damaged cells would be more accessible to these molecules.

Cumulatively, these results are consistent with the notion that persistent P. aeruginosa bacteria present in aggregated structures, which are dispersible upon addition of anti-biofilm molecules, contribute to the adaptive antibiotic-resistant phenotype. Moreover, these findings validate the pertinence of our in vivo model to identify compounds that could potentiate the effects of conventional antibiotics.

Ethics statement for zebrafish

All zebrafish experiments described in the present study were conducted at the University of Montpellier by following the 3rs -Replacement, Reduction and Refinement- principles according to the European Union guidelines for handling of laboratory animals (https://environment.ec.europa.eu/topics/chemicals/animals-science_en) and were approved by the “Direction Sanitaire et Vétérinaire de l’Hérault” and the “Comité d’Ethique pour l’utilisation d’animaux à des fins scientifiques” under reference CEEA-LR-B4-172-37. Breeding of adult fish adhered to the international guidelines specified by the EU Animal Protection Directive 2010/63/EU. All experiments were performed before the embryos free feeding stage (5 dpf) and did not fall under animal experimentation law according to the EU Animal Protection Directive 2010/63/EU. Embryos were euthanized using the anesthetic tricaine up to a lethal dose before bleach treatment. Embryo manipulation, handling, and euthanasia were performed by well-trained and authorized staff. Embryos were euthanized using an anesthetic overdose of buffered tricaine before bleach treatment.

Clinical strains used in the study and growth conditions

P. aeruginosa clinical strains were routinely grown at 37 °C in lysogeny broth (LB) with shaking at 180 rpm. Clinical strains A6520, B6513 and C6490 were isolated at the teaching hospital of Besançon (France) from three individual cystic fibrosis patients aged 4, 1 and 17 years, respectively showing a transient (A6520, B6513) or chronic (C6490) lung colonization with P. aeruginosa. As recommended by international guidelines, all these cystic fibrosis patients were repeatedly treated with combinations of antibiotics to control or attempt to eradicate their pulmonary infection (individual treatment regimens were not available). Genome sequencing was carried out to classify the strains. Sequence Types (STs) were determined according to the MLST scheme available at PubMLST (https://pubmlst.org). Strains A6520, B6513 and C6490 were found to belong to the same phylum as reference strain PAO1 [58], to harbor the exotoxin-encoding gene exoS, and to be genotypically distinct (ST274, ST27 and ST633, respectively). This Whole Genome Shotgun project (whole genomes sequences of the clinical isolates) has been deposited at DDBJ/ENA/GenBank under the accession JAWDIG000000000 (A6520), JAWDIH000000000 (B6513) and JAWDII000000000 (C6490). RP73 is a well described CF P. aeruginosa [26].

Chemicals

All antibiotics were purchased from Sigma-Aldrich. Ciprofloxacin was dissolved at 25 mg/mL in 0.1 M hydrochloric acid. Ofloxacin was dissolved at 20 mg/mL in 1 M NaOH. Colistin sulfate salt and Tobramycin were dissolved at 8 mg/mL in ultrapure water. The human Atrial Natriuretic Peptide (hANP) was purchased from Tocris Bioscience (Bio-Techne) and was dissolved at 1 mg/mL in ultrapure water and stored at −20 °C. The cis-2-Decenoic acid (CDA) was purchased from Sigma-Aldrich and was dissolved at 5.8 mM in 100% DMSO and stored at −20°C.

Minimum inhibitory concentration (MIC) assays

Susceptibility levels of strain B6513 to the fluoroquinolones ciprofloxacin (0.5 µg/mL) and ofloxacin (0.5 µg/mL), the polymyxin colistin (4 µg/mL), and the aminoglycoside tobramycin (1 µg/mL) were determined by the standard microdilution method in 96-well plates. Overnight cultures in LB broth were diluted in the same medium to an initial inoculum of OD_600nm = 0.1. MIC was defined as the lowest concentration that inhibited bacterial growth.

Construction of stable fluorescent P. aeruginosa strains

Strains with chromosomally encoded GFP were obtained by triparental mating by using a recombinant integrative plasmid miniCTX carrying the PX2-GFP fusion [59] (from Ina Attree, Université Grenoble Alpes, France). Following overnight growth in LB broth with appropriate antibiotics, E. coli TOP10 carrying either helper plasmid pRK2013 [35] or miniCTX-PX2-GFP were mixed (20 µL of each culture) and deposited as drops onto the surface of LB plates for 2 h at 37 °C. In parallel, the overnight cultures of P. aeruginosa strains were incubated at 42 °C. 20 µL fractions of these stationary phase cultures were then individually added to the E. coli dry drops and left at 37 °C for 5 h. The bacterial spots were then scrapped off and directly plated on LB plates containing Irgasan (25 µg/mL) and tetracycline (from 50 to 200 µg/mL) to select P. aeruginosa transconjugants. After 18 h at 37 ^oC, several colonies were streaked out on LB medium to verify the stable GFP expression, and an individual GFP⁺ colony was cultured in liquid LB to obtain a glycerol stock.

Zebrafish lines

Wild-type AB and Golden lines or AB/Golden mixed genetic background zebrafish were used for survival and CFU experiments. For live imaging, lines carrying fluorescent macrophages Tg(mfap4:mCherry-F)ump6TG [60], neutrophils Tg(LysC:dsRed)nz50 [61] or mesenchymal cells Tg(rcn3:Gal4/UAS:mCherry) [62,63] were used. Fish maintenance, staging and husbandry were performed as described [64]. Eggs were obtained by natural spawning, collected in petri dishes and incubated at 28°C in fish water composed of 60 μg/mL sea salts (Instant Ocean) in distilled water supplemented with 0.4 mM NaOH and 0.1% methylene blue. Embryos and larvae were staged according to [65]. For experiments, larvae were used at 2-day post-fertilization (dpf) until 5 dpf.

Infection by immersion of injured zebrafish embryos

The protocol was essentially as described earlier [23], with few modifications. Overnight bacterial cultures were diluted at 1:20 in fresh LB broth and incubated until the OD_600nm reached approx. 0.8–1. Cultures were centrifuged at 4000 rpm for 10 min and re-suspended in fish water at approx. 10⁷ bacterial/mL. The bacterial load was determined by subsequent plating onto LB agar after dilution into phosphate-buffered saline (PBS). 2 dpf embryos, previously dechorionated, were anesthetized with tricaine (300 µg/mL, Sigma-Aldrich) and injured at the tail fin using 25-gauge needles under a stereomicroscope (Motic). Embryos were distributed into a 6-well plate containing 4 mL of bacterial suspension (or fish water as a control) immediately after the injury, and incubated at 28 °C for 1.5 h. Two washes in fish water were subsequently performed (30 min in 10 mL and a few minutes in 4 mL, respectively) to eliminate bacteria in the bath. Finally, larvae were transferred individually into a 96-well plate (survival experiment) or a 24-well plate (CFU measurement and microscopy) containing fish water, and incubated using a light/dark cycle at 28 °C. For survival curves, death was determined based on the absence of heart beat after visual inspection.

Bacterial load measurement in infected embryos

Before CFU quantification, infected embryos were washed for a few minutes in 4 mL of fish water and subsequently crushed individually using a pestle in 100 µL of PBS. Then 100 µL of Triton X-100 (1% final concentration) were added for 10 min to liberate bacteria from residual cells/tissues. Following lysis, several dilutions in PBS were spotted on LB agar plates and incubated approx. 18 h at 37 °C. Only GFP⁺ colonies were considered for counting. For CFU measurement following treatment, embryos were washed before the treatment and not prior the crushing (see below). The number (n) indicated in figure legends reflects the number of independent experiments.

Embryo imaging

After reaching the 50% epiboly stage, embryos were maintained in fish water supplemented with the melanization inhibitor Phenylthiourea (PTU) to prevent pigmentation. Before imaging, larvae were anesthetized with 300 µg/mL tricaine and placed in 35 mm glass-bottom dishes (FluoroDish, World Precision Instruments) and immobilized with 1% low-melting-point agarose (Sigma-Aldrich), covered with fish water after solidification. Zebrafish larvae were imaged at indicated hpi using the ANDOR CSU-W1 confocal spinning disk on an inverted NIKON microscope (Ti Eclipse) with ANDOR Zyla 4.2 sCMOS camera (40x water/NA 1.15 objective). Optimal laser power and exposure times were determined for each strains (GFP) and cell types (mCherry), and were identical between independent experiments. Settings of brightness and contrast used for a better visualization were the same between all independent experiments. The 3D files generated by multi-scan acquisitions were processed using Image J and compressed into maximum intensity projections. Brightness and contrast were adjusted for better visualization. Deconvolution of images was done to improve resolution with Huygens (SVI).

Quantification of macrophage recruitment, bacterial volumes and percentage of intra-macrophage GFP⁺ foci

As a proxy to quantify macrophage recruitment at the injury site following spinning disk confocal imaging, the area formed by positive pixels (mCherry⁺) was measured at the median plan of each stack. To that aim, using Fiji, a threshold of positivity was applied using the ‘’IsoData” method. 3D reconstructions and volume quantification of GFP⁺ events were performed using the Imaris ‘’surface” tool. Following the generation of GFP⁺ 3D objects, very small events (volume <0.5 μm³) with non-rod shapes were considered as autofluorescence and eliminated. In addition, rare GFP⁺ objects localized outside the tissues (e.g., at the surface of the fish) were manually removed from the analysis.

Quantification of the percentage of GFP⁺ foci located inside macrophages over time was performed on embryos with at least one event of intra-macrophage foci (7 to 11 embryos at each time point). Confocal Z-stacks were converted into maximal intensity projections (MIP) in the GFP channel and a threshold was applied on the 2D images at a pixel value two times higher than the background value. Using a composite GFP/mCherry Z-stack in parallel, GFP⁺ foci identified on the MIP were localized outside or inside macrophages and grouped in the ROI manager to measure their surface. A global ‘’GFP area” was calculated before deducing the percentage of intra-macrophage surface.

Treatment of infected embryos with antibiotics and anti-biofilm molecules, and assessment of antibiotic resistance of bacteria recovered from antibiotic treated animal

Prior any treatment, embryos were systematically washed for a few minutes in 4 ml of zebrafish water. Infected larvae were incubated with 40 x MIC of antibiotics, except for colistin (2.5 x MIC). Depending of the antibiotic, controls were done with fish water alone or supplemented with NaOH or HCl at the proper concentration. At least 5 embryos of the same condition were treated together for 30 min in 1 mL at room temperature. For CDA (20 µM) and hANP (10 µM), treatments were applied individually for 5 h at 43 hpi in 200 µL at 28 °C. Control conditions were fish water alone (hANP) or supplemented with 0.34% DMSO (CDA). Independent experiments were systemically performed 3 times, with at least 5 larvae per condition (i.e., minimum 15 embryos in total).

Bacteria recovered from antibiotic treated animals at 48 or 72 hpi were challenged against antibiotics in vitro to assess if bacteria acquired antibiotic resistance. Bacteria recovered from CFU plating were grown O/N in LB medium in parallel with an in vitro colony (not obtained from zebrafish infection). Bacterial cultures were treated with ciprofloxacin (20 µg/ml) or tobramycin (20 µg/ml) for 30 min and the susceptibility to antibiotics was determined by surviving CFU. Bacteria issued from embryos (3 independent colonies tested) were as sensitive as the original B6513.

Cutaneous infection in mice

Animal experiments were performed in accordance with the Canadian Council on Animal Care (CCAC) guidelines and were approved by the University of British Columbia Animal Care Committee protocol (A23-0030). Mice used in this study were outbred CD-1 mice (female). All animals were purchased from Charles River Laboratories, Inc. (Wilmington, MA, United States) and were 7–8 weeks of age at the time of experiments. Mice weighed 25 ± 2 g at the experimental start point and standard animal husbandry protocols were employed.

We tested the virulence of P. aeruginosa clinical isolates (A6520, B6513, and C6490) and the Liverpool Epidemic Strain LESB58 in a nuanced model of cutaneous high-density infection as previously described [16] with minor modifications. All strains were sub-cultured at 37°C with shaking (250 rpm) to an OD_600nm = 1.0 in LB. Cells were washed twice with sterile phosphate buffered saline (PBS) and resuspended to a final OD_600nm = 0.5 or 1.0 for clinical isolates or LESB58 strains, respectively. Strains were used to form high-density abscess infections to model invasive or chronic infections. Abscesses were formed by injection of 50 μL of bacteria on the left dorsum of mice for 3 or 7 days. Disease progression and overall welfare of animals was monitored daily up to day three, and weekly thereafter. At experimental endpoint, animals were euthanized using carbon dioxide followed by cervical dislocation, and abscess lesion size was measured using a caliper. Abscesses were harvested in PBS and homogenized using a Mini-Beadbeater (BioSpec Products, Bartlesville, OK, United States) for bacterial enumeration on LB. Two independent experiments containing 2–4 biological replicates each were performed.

Electron microscopy

Samples were chemically fixed using 2.5% glutaraldehyde (Electron Microscopy Sciences # 16216) in fish water. After fixation, embryos were stored at 4 °C in the fixation solution until subsequent processing. The procedure for embedding in resin was adapted from a previous method [66]. All the procedure, except the overnight incubation in uranyl acetate, was performed using a Pelco Biowave PRO+ Microwave processing systems (TED Pella) following the program indicated in Table S1. Samples were post-fixed with 2% osmium tetroxide (O_sO₄) in 0.1M cacodylate buffer pH 7.4 containing 5 mM CaCl₂, immediately followed by 1.5% K-ferrocyanide in the same buffer. After washing, samples were treated with 1% thiocarbohydrazide at 60 °C, washed with distilled water before a second incubation in 2% OsO_4. After washing, samples were then incubated overnight in 1% uranyl acetate at 4 °C. Samples were next heated at 40°C and further processed in the microwave (Table S1), washed and incubated in lead aspartate pre-heated at 50°C. Dehydration was performed with growing concentrations of acetonitrile. Samples were then impregnated in Epon Hard+(TM) resin, and polymerized 48 h at 60 °C. All chemicals were from EMS.

Thin serial sections were made using an UCT ultramicrotome (Leica) equipped with a Jumbo ultra 35° diamond knife (Diatome). Section ribbons were collected on silicon wafers (Ted Pella) with the help of an ASH2 manipulator (RMC Boeckler). Sections were imaged with a Zeiss Gemini 360 scanning electron microscope on the MRI EM4B platform under high vacuum at 1.5 kV. Final images were acquired using the Sense BSD detector (Zeiss) at a working distance between 3.5 and 4 mm. Mosaics were acquired with a pixel size of 5 nm and a dwell time of 3.2 µs.

Macrophage infection and quantification of intracellular bacteria

J774 cells (murine macrophage cell line) were maintained at 37 °C in 5% CO₂ in Dulbecco’s modified Eagle medium (DMEM, Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Gibco). The infection of J774 macrophages by P. aeruginosa was carried out essentially as described previously [35]. J774 macrophages (5 x 10⁵ cells/well) were infected by mid-log phase bacteria in PBS, at a MOI of approx. 10. Infection synchronization was done by a 5 min centrifugation at 1000 rpm of the 24-well plate, and bacterial phagocytosis was allowed to proceed for 25 min. Cells were then washed three times with sterile PBS and fresh DMEM medium supplemented with 400 µg/mL gentamicin was added, which was retained throughout the infection to kill non-phagocytosed bacteria. Macrophages were lysed after 20 min (T0) or 2 h (T1) of gentamicin treatment, by using 0.1% Triton X-100 and the number of viable bacteria was determined by subsequent plating onto LB agar plates. Survival rate at T1 was compared to the number of internalized bacteria at T0. The cytotoxicity driven by internalized bacteria was assessed using Trypan blue as described [34].

Live microscopy on cultured macrophages

J774 macrophages were seeded in glass bottom 8 wells μ-slide (Ibidi #80827) in DMEM medium supplemented with 10% FBS and infected with P. aeruginosa strains expressing GFP as described above. Imaging started after 30 min of phagocytosis, when the media was changed to DMEM supplemented with 400 μg/ml gentamicin until 3 h post phagocytosis. Cells were imaged using an inverted epifluorescence microscope (AxioObserver, Zeiss), equipped with an incubation chamber set-up at 37 °C and 5% CO₂ and a AxioCam 503 Mono camera (Photometrics). Time-lapse experiments were performed, by automatic acquisition of random fields using a 63X Apochromat objective (NA 1.4). The frequency of acquisition is indicated in figure legends. Image treatment and analysis were performed using Image J.

Biofilm formation

Biofilm formation was assessed in low magnesium medium in glass tubes 24 h at 30 °C under static condition as described previously [67]. After 24h, bacterial growth was measured by OD_600nm and tubes were carefully washed with water. The biofilm at the air-liquid interphase was stained using crystal violet (CV) 0.1% during 15 min at room temperature. After staining, tubes were washed with water and CV was extracted using acetic acid (30%) and quantified by measuring the OD_570nm.

Swarming motility

A modified form of basal medium (BM2) consisting of 62 mM potassium phosphate buffer (pH = 7.0), 2 mM MgSO₄, 10 μM FeSO₄, 20 mM glucose and 0.1% casamino acids (CAA) was used for swarming assays. Subcultures were adjusted to a starting OD₆₀₀ = 0.1 in BM2 and grown to an OD_600nm = 0.4–0.6 for spot inoculation on solidified medium (0.4% agar). Plates were incubated for 18–24 h at 37 °C, were imaged with a BioRad ChemiDoc and surface area coverage of the plate was measured in ImageJ2 software (v2.14.0). Three biological replicates were examined.

Pyocyanin production

The concentration of pyocyanin in bacterial supernatants was determined spectrophotometrically after extraction with chloroform and 0.2 M HCl, as described elsewhere [68]. Absorbance at 520 nm was read using a Biotek Synergy H1 Microplate Reader. Three biological replicates were examined.

In vitro efficiency of hANP

The flow cell system, which allows for continuous bacterial biofilm formation, is assembled, prepared and sterilized as described earlier [69]. For studying the impact of hANP peptide on established strain B6513 biofilm, we used an established protocol [39]. Briefly, bacterial cells from an over-night culture, were recovered by centrifugation (10 min, 7,500 rpm) and washed with sterile physiological water (0.9% NaCl). Each channel of the flow cell (1 mm x 4 mm x 40 mm, Bio centrum, DTU) was inoculated with 300 μL of bacterial suspension prepared at an optical density of OD_580nm=0.1. Bacterial adhesion was allowed without any flow for 2 h at 37 °C. After 2 h of adhesion, the LB medium was pumped with a flow rate of 2.5 mL/h at 37 °C for 24 h. Next, the 24 h-old biofilm was exposed for 2 h to 300 µL of hANP (0.1 µM) or 300 µL of ultra-pure distilled water (control condition), added to each channel of the flow cell and without flow. Prior to image acquisition, biofilm cultures were then rinsed with LB medium using a 2.5 mL/h flow rate for 15 min. Finally, bacterial cells were marked with 5 µM of SYTO9 green-fluorescent dye (Invitrogen) and observed using confocal laser scanning microscopy (CLSM). CLSM observations of biofilms were performed using a Zeiss LSM710 microscope (Carl Zeiss Microscopy) using a x40 oil immersion objective. In order to capture the entire biofilm depth, images were taken every millimeter. For visualization and processing of three-dimensional (3D) images, the Zen 2.1 SP1 software (Carl Zeiss Microscopy) was used. Using the COMSTAT software (http://www.imageanalysis.dk/), quantitative analyses of image stacks were carried out [70].

Statistical analysis

GraphPad Prism 8.3.0 was used to perform all statistical tests and create graphs. The box plots represent the median, as well as min and max values with the interquartile range. The indicated test used to analyze each dataset was chosen depending on the normality of the data. Multiple comparisons were done by one-way ANOVA or Kruskall-Wallis test, followed by Tukey’s or Dunn’s pairwise comparison, respectively. The Dunnett’s many-to-one comparison was used following a one-way ANOVA. Mann-Whitney U test was used to compare two groups.