An In Vitro Model of Nonattached Biofilm-Like Bacterial Aggregates Based on Magnetic Levitation

An increasing amount of evidence indicates that chronic infections are associated with nonattached biofilm-like aggregates formed by pathogenic bacteria. These aggregates differ from biofilms because they form under low-shear conditions within the volume of biological fluids and they do not attach to surfaces. Here, we describe an in vitro model that provides nonattached aggregate formation within the liquid volume due to magnetic levitation. Using this model, we demonstrated that despite morphological and functional similarities of nonattached aggregates and biofilms, strains that exhibit good biofilm formation might exhibit poor nonattached aggregate formation, suggesting that mechanisms underlying the formation of biofilms and nonattached aggregates are not identical. The magnetic levitation approach can be useful for in vitro studies of nonattached aggregate formation and simulation of bacterial behavior in chronic infections.

ABSTRACT

Chronic infections are associated with the formation of nonattached biofilm-like aggregates. In vitro models of surface-attached biofilms do not always accurately mimic these processes. Here, we tested a new approach to create in vitro nonattached bacterial aggregates using the principle of magnetic levitation of biological objects placed into a magnetic field gradient. Bacteria grown under magnetic levitation conditions formed nonattached aggregates that were studied with confocal laser scanning microscopy (CLSM) and scanning electron microscopy (SEM) and characterized quantitatively. Nonattached aggregates consisted of bacteria submerged into an extracellular matrix and demonstrated features characteristic of biofilms, such as a polymeric matrix that binds Ruby Red and Congo red dyes, a prerequisite of bacterial growth, and increased resistance to gentamicin. Three quantitative parameters were explored to characterize strain-specific potential to form nonattached aggregates: geometric sizes, relative quantities of aggregated and free-swimming bacteria, and Congo red binding. Among three tested Escherichia coli strains, one strain formed nonattached aggregates poorly, and for this strain, all three of the considered parameters were different from those of the other two strains (P < 0.05). Further, we characterized biofilm formation on plastic and agar surfaces by these strains and found that good biofilm formation ability does not necessarily indicate good nonattached aggregate formation ability, and vice versa. The model and quantitative methods can be applied for in vitro studies of nonattached aggregates and modeling bacterial behavior in chronic infections, as it is important to increase our understanding of the role that nonattached bacterial aggregates play in the pathogenesis of chronic diseases.

IMPORTANCE An increasing amount of evidence indicates that chronic infections are associated with nonattached biofilm-like aggregates formed by pathogenic bacteria. These aggregates differ from biofilms because they form under low-shear conditions within the volume of biological fluids and they do not attach to surfaces. Here, we describe an in vitro model that provides nonattached aggregate formation within the liquid volume due to magnetic levitation. Using this model, we demonstrated that despite morphological and functional similarities of nonattached aggregates and biofilms, strains that exhibit good biofilm formation might exhibit poor nonattached aggregate formation, suggesting that mechanisms underlying the formation of biofilms and nonattached aggregates are not identical. The magnetic levitation approach can be useful for in vitro studies of nonattached aggregate formation and simulation of bacterial behavior in chronic infections.

INTRODUCTION

It is generally believed that bacteria exist in two states, single floating planktonic cells and surface-associated cell aggregates known as biofilms (1). Bacterial biofilms are surface-associated three-dimensional (3D) multilayer structures formed by bacteria and self-produced matrix consisting of exopolysaccharide, proteins, extracellular DNA, and lipids (2, 3). Protection mechanisms provided by biofilms, such as limited diffusion of some antibiotics, accumulation of antibiotic-modifying enzymes in the matrix, and persister cells, provide hundreds-fold increased resistance to antibiotics and other microbicidal treatments compared to single-cell planktonic bacteria (4–11). Biofilms formed by pathogenic bacteria play an important role in clinical pathology. In addition to their well-established role in the development of hospital-acquired infections associated with implants, catheters, and other biomedical devices, biofilms formed within human organs and tissues are involved in the establishment of chronic infections, including chronic wounds, bronchitis, arthritis, otitis, and rhinosinusitis (12–19). To date, biofilms remain a serious challenge to chronic infection therapy.

Biofilms are typically described as structures that form on solid surfaces and less frequently on liquid surfaces (1, 20–22). These solid surface-associated biofilms are best studied using multiple in vitro models that have played a key role in understanding the main stages of biofilm growth and maturation and the mechanisms by which they are controlled. The most popular biofilm models include microtiter plate systems and constant flow fermenters (23–31).

The development of microscopic techniques allowed biofilm analysis ex vivo and in vivo in samples derived from mucosal and epithelial tissues, joints, and chronic wounds. In many cases, chronic disease-associated biofilms were not directly associated with human tissue surfaces but, rather, levitated within the volume of physiological liquids. Such nonattached aggregates formed by bacteria embedded into polymeric matrix have been described for a wide range of pathogenic bacteria, including toxigenic Escherichia coli, Pseudomonas aeruginosa, Achromobacter spp., Mycobacterium abscessus, Borrelia spp., Actinobacillus pleuropneumoniae, and Staphylococcus aureus (32–42).

The absence of an external supporting surface in aggregate formation is an important difference that points to limitations of existing in vitro biofilm models (reviewed by Roberts and coauthors [43]). Another specific feature of clinically relevant bacterial aggregates is the absence of the shear stress in the host tissues (4, 16, 35, 39, 41, 42, 44). To study properties of nonattached aggregates, an increasing number of in vitro models has appeared in recent years. Monitoring of bacterial aggregation in nonshaking cultures in highly viscous media, within agar blocks, or in the presence of physiological liquids demonstrated that nonattached aggregates share some important characteristics with surface-attached biofilms, including the presence of extracellular matrix, requirement for bacterial multiplication, and increased resistance to antimicrobial treatments (45–48). On the other hand, features, such as the requirement for bacterial motility, matrix composition, and regulatory mechanisms might differ between nonattached aggregates and classic biofilms (32, 40, 49). The major limitation for in vitro models of nonattached aggregates is sedimentation of aggregates via gravitational force even in highly viscous media that restricts the observation time.

Here, we suggest an alternative approach that facilitates the formation of nonattached nonsedimenting bacterial aggregates under magnetic levitation conditions. Magnetic levitation is widely used in industry and research to create conditions where an object is suspended with no support other than magnetic fields that counteract gravitational effects (50). To achieve an effect, a combination of permanent magnets and diamagnets or superconductors is typically used. In biology, magnetic levitation is used in microfluidic studies for the development of systems for cell collection and analysis (51–53). Here, we used a recently described magnetic levitation system that was developed to provide scaffold-free biofabrication of tissue spheroids (54, 55). We applied this magnetic levitation system to study nonattached aggregates formed by three Escherichia coli strains. The described techniques of nonsurface attached aggregate creation together with the suggested quantitative methods can be used in future studies of surface-independent bacterial aggregates.

RESULTS

Bacterial behavior under magnetic levitation conditions.

Here, we used a previously described system that provides magnetic levitation of biological objects, such as mammalian cells (54, 55). To obtain a levitation effect, cells were suspended in gadobutrol solution and placed into the magnetic field gradient. The magnetic field gradient is created in a specifically designed device named a magnetic bioprinter (Fig. 1). Gadobutrol (Gd-DO3A-butrol) is a chemical derivative of the rare-earth element gadolinium, which possesses paramagnetic properties (56, 57). Gadobutrol is FDA approved for tissue contrast enhancement in magnetic resonance imaging (MRI) (58, 59). Preclinical and clinical studies demonstrated that gadobutrol has low toxicity for mammals and humans (60–62).

FIG 1.

Parts of the magnetic trap. (A) The magnetic bioprinter scheme. (1) Case; (2) magnetic unit; (3) peephole; (4) portal for installing a cuvette. (B) Direction of cuvette placement into the bioprinter. (C) Combination of permanent magnets that provide the magnetic field gradient. (1) Two magnetic rings combined by poles of the same polarity; (2) metal box; (3) fixation of the box with cover. (D) The internal part of the bioprinter where the cuvette is placed. (E) The magnetic field formed within the bioprinter.

The described system was tested with E. coli strains. E. coli is a standard model for biofilm studies. Moreover, pathogenic E. coli forms nonattached aggregates in urinary tract infection (34). To test gadobutrol toxicity to bacteria, we grew E. coli in LB broth supplemented with 20% gadobutrol (vol/vol) or with 20% phosphate-buffered saline (PBS) (vol/vol). No toxic effect on bacteria was noted after 24 h of growth (see Fig. S1 in the supplemental material).

The toxigenic E. coli ATCC 43890 culture was grown overnight in LB broth supplemented with 0.2 M gadobutrol and placed into the bioprinter (Fig. 2). The special window allowed monitoring of bacterial behavior in the bioprinter in real time. During the 1st hour of observation, no macroscopic changes in bacterial distribution over the volume were noted (Fig. 2A). Starting from the 2nd hour, clearance of peripheral regions was observed. Further observations demonstrated that the clearance seemed to be attributed to slow bacterial movement from the periphery to the center. Nine hours after the experiment started, peripheral regions were transparent, whereas the center of the cuvette was occupied by bacteria (Fig. 2A). Twenty-four hours after the experiment started, the bacterial conglomerate diminished in volume compared to the 9 h time point but still levitated within the tube. However, outlying regions seemed transparent. The diameter and the height of the conglomerate were approximately 4 and 10 mm, respectively. Prolonged incubation did not cause further changes, suggesting that forces acting on bacteria were in equilibrium. After the flask was removed from the bioprinter, bacteria were distributed throughout the flask.

FIG 2.

Bacterial behavior under magnetic levitation conditions. (A) Overnight E. coli ATCC 43890 culture grown in the LB broth was supplemented with 20% (vol/vol) of 1 M gadobutrol to obtain a final gadobutrol concentration of 0.2 M and was placed within the bioprinter. Images were obtained through a special peephole (see Fig. 1) immediately (0 h) and at 9 h and 24 h, and the width and height of the bacterial population were measured at the same time points. (B) Overnight E. coli ATCC 43890 culture was supplemented with 10% (vol/vol) of 1 M gadobutrol to obtain the final gadobutrol concentration of 0.1 M and was placed within the bioprinter. The image was obtained at 24 h.

The same experiment was performed with 0.1 M gadobutrol. When a lower concentration of gadobutrol was used for the process, similar results were obtained. However, the bacteria occupied a larger portion of the volume with a diameter of approximately 8 mm, and bacteria were located closer to the bottom of the tube than when experiments were performed with 0.2 M gadobutrol (Fig. 2B). Upon removal from the bioprinter, bacteria were uniformly distributed by volume.

The redistribution of the bacterial suspension over the volume after removal of the magnetic field suggests that the observed conglomerate was not the true aggregate but, rather, a diameter of the magnetic trap where the gradient of the magnetic field was sufficiently high to hold bacteria. The specific behavior of diamagnets and paramagnets in the magnetic field gradient that underlined this effect is discussed below.

Bacterial growth under magnetic levitation conditions.

Next, we changed the experimental conditions to allow bacterial growth. The E. coli ATCC 43890 overnight bacterial culture was diluted 1:100 with fresh LB broth containing 0.2 M gadobutrol and incubated in the magnetic bioprinter at 37°C for 24 h. At the end of the experiment, a small bacterial aggregate levitated in the center of the tube, whereas peripheral regions were transparent. The size of the aggregate was noticeably smaller than the size of the conglomerate formed by the overnight culture (compare Fig. 3A and 2A). After the flask was removed from the bioprinter, the aggregate slowly settled. In contrast to the experiments using a nongrowing overnight culture, bacteria were not uniformly distributed throughout the volume after the flask was removed from the bioprinter. These results suggested that bacterial growth was necessary for the bacteria to adhere together and aggregate. Reducing the gadobutrol concentration down to 0.1 M resulted in the appearance of the aggregate near the bottom. This aggregate settled down at 24 hours postinoculation (hpi) and was observed as a spot on the bottom of the syringe (Fig. 3B). LB broth containing 0.2 M gadobutrol (hereafter referred to as the paramagnetic LB broth) was used in all subsequent experiments.

FIG 3.

Bacterial growth under conditions of magnetic levitation. (A) Overnight E. coli ATCC 43890 culture was diluted 1:100 with 0.2 M paramagnetic LB medium (LB supplemented with 0.2 M gadobutrol) and incubated within the bioprinter at 37°C for 24 h; the image was obtained after the cuvette was removed from the bioprinter. (B) Overnight E. coli ATCC 43890 culture was diluted 1:100 with LB supplemented with 0.1 M gadobutrol and incubated within the bioprinter at 37°C for 24 h. The aggregate settled to the bottom and was laid on the wall of the syringe. Arrows point to bacterial aggregates.

Bacteria in aggregates are viable and can leave the magnetic trap.

To quantify viable bacteria in the aggregate, the aggregate was resuspended in 1 ml of PBS after the medium was carefully removed. The aggregate included (1.47 ± 0.47) × 10⁸ CFU ml⁻¹. To determine whether bacteria could leave the aggregate, the optically clear medium from the side of the aggregate was taken, and serial dilutions were plated to quantify bacterial loads. Although the medium seemed transparent, the concentration of bacteria outside the central aggregate reached (0.007 ± 0.001) × 10⁸ CFU ml⁻¹. Therefore, approximately 0.5% of the total bacteria were free floating, while most bacteria were located in the aggregate. No noticeable difference was noted between results obtained when the entire suspension from the trap was plated and when the pure aggregate was plated. This finding is consistent with the relatively low percentage of free-swimming bacteria (Fig. 4). Further, we prolonged incubation to 6 days (144 h). The total number of viable bacteria in the aggregate increased less than 2-fold and reached (2.4 ± 0.4) × 10⁸ CFU, whereas the number of free-swimming bacteria was (0.1 ± 0.4) × 10⁸ CFU, representing approximately 4.5% of the total population (Fig. 4).

FIG 4.

Bacterial viability in the magnetic trap. ATCC 43890 bacteria were grown under conditions of magnetic levitation for 24 h or 144 h. Then, the medium outside the aggregate was carefully removed, and the aggregate itself was suspended in sterile PBS. Serial dilutions from the aggregate suspension and the medium outside the aggregate were plated to count aggregated and free-swimming bacteria. Data represent the mean ± SD from three independent experiments.

Microscopic assessment of levitating bacterial aggregates revealed morphological features characteristic of biofilms.

To assess aggregate morphology, E. coli ATCC 43890 grown under conditions of magnetic levitation for 144 h was fixed with glutaraldehyde and retrieved from the tube. The macroscopic aggregate fell apart in small pieces. The largest of these pieces was approximately 0.2 mm, whereas the majority were smaller (Fig. S2). These pieces were subjected to microscopic studies (Fig. 5). The pieces obtained from the 24-h aggregate were too small to manipulate.

FIG 5.

Microscopic view of aggregates formed by E. coli ATCC 43890. Bacteria were grown under conditions of magnetic levitation for 7 days and fixed with 2.5% glutaraldehyde. Then, a portion of a sample was stained with SYBR green and FilmTracer SYPRO Ruby biofilm matrix stain (A and B), and another portion was prepared for SEM. (A) CLSM demonstrated that bacterial aggregates are formed by bacteria and noncellular matrix; elongated and normal bacterial cells are observed. (B) 3D reconstruction of the sample. (C) SEM-supported view of aggregates formed by bacteria (arrows) and the matrix with a vesicular structure (arrowhead).

The fixed sample was labeled with nucleic acid binding fluorescent SYBR green dye and FilmTracer SYPRO Ruby biofilm matrix stain and studied with confocal laser scanning microscopy (CLSM). SYBR green, which preferentially stained nuclear DNA, revealed multiple bacterial cells (Fig. 5A). Bacterial cells were submerged into matrix stained with Ruby biofilm matrix stain to form a 3D structure (Fig. 5B). To better characterize the morphology of the bacterial aggregates formed under magnetic levitation, E. coli ATCC 43890 was studied with scanning electron microscopy (SEM). The observed 3D structures supported the observation that aggregates are formed by bacterial cells and matrix. The structure of the matrix was rather vesicular and bubbly. Cell morphology CLSM analysis revealed elongated cells that reached up to 4- or 5-fold the length of a normal cell. Consistent with CLSM results, SEM analysis revealed long bacterial cells. Moreover, short and almost oval cells were observed with SEM.

Overall, microscopic studies demonstrated that the aggregate formed by E. coli ATCC 43890 consisted of bacteria submerged into the polymeric matrix interacting with the standard biofilm matrix tracer Ruby Red. Here, aggregate morphology resembled that of surface-attached biofilms.

Nonattached aggregates exhibited increased antibiotic resistance.

Increased resistance to antibiotics is a hallmark of bacterial biofilms. Introduction of 80 μg ml⁻¹ gentamicin (8-fold MIC) to the levitating ATCC 43890 aggregate caused only a 5-fold decrease in the number of live bacteria 24 h later (Fig. 6). This concentration completely restricted ATCC 43890 growth in planktonic cultures in the standard test.

FIG 6.

Aggregates improved bacterial resistance to antibiotics. Gentamicin was added to ATCC 43890 aggregates up to 8 MIC for 24 h, and bacteria from the gentamicin and control cultures were plated after an additional 24 h. Results are shown as the percentage of live bacteria relative to the control culture.

Comparative study of three E. coli strains revealed strain-specific differences in bacterial behavior under magnetic levitation conditions.

In parallel with ATCC 43890, which is a toxigenic enterohemorrhagic strain, we tested two additional E. coli strains, the laboratory strain JM109, which is commonly employed in bioengineering applications, and the probiotic strain M17. All three strains formed aggregates when grown under magnetic levitation conditions (Fig. 7). Geometric parameters of the aggregate formed by the strain M17 were similar to those of ATCC 43890. Geometric parameters of the JM109 aggregate were noticeably larger (Fig. 7A). The percentage of bacteria in the aggregate relative to the total population of aggregate-bound and free-swimming bacteria was noticeably reduced for JM109 compared with other strains (P < 0.05; Fig. 7B). Further, fixation of the JM109 aggregate did not increase the appearance of stable visible macroscopic pieces (data not shown). We suggested that JM109 formed nonattached aggregates less efficiently than the two other strains. In particular, we suggested that JM109 might be deficient in production of the matrix that fastened bacteria to form a dense aggregate. To obtain more quantitative data confirming this suggestion, we applied a modification of Congo red staining to the bacterial cultures growing under magnetic levitation. Congo red dye, which was first described and used as an amyloid-binding dye in humans, is widely used in microbiology to visualize biofilm matrix components (63–65). Congo red is nontoxic, does not inhibit bacterial growth, and can be added directly into the growth medium. Overnight cultures of ATC43980 or JM109 were diluted 1:100 in the paramagnetic growth medium supplemented with Congo red (the paramagnetic LB/Congo red [CR] medium) and allowed to grow. The formed aggregate was treated as described in Materials and Methods, and bound dye was washed from the aggregate with 50% ethanol. The aggregate formed by the strain ATCC 43890 bound at least 2-fold more CR dye than JM109 (Fig. 7C; P < 0.01). These data suggest that JM109 aggregates included less matrix and are consistent with other data indicating that the JM109 strain less efficiently forms nonattached aggregates.

FIG 7.

Comparative study of three E. coli strains grown under magnetic levitation conditions. The E. coli strains JM109, ATCC 43890, and M17 were grown under conditions of magnetic levitation as described in Fig. 3. (A) Geometric parameters of the aggregates. (B) Percentage of bacteria bound in aggregates and free-swimming bacteria. (C) Matrix formation by bacterial aggregates grown under magnetic levitation conditions. ATCC 43890 and JM109 were grown under conditions of magnetic levitation in paramagnetic LB broth supplemented with Congo red for 144 h. Congo red binding by bacterial aggregates was determined as described in the text. Data are from at least three independent experiments. *, P < 0.05; **, P < 0.01.

Nonattached aggregates and biofilms form via nonidentical mechanisms.

The strain JM109 is known to form biofilms on solid surfaces (66). To assess differences in the biofilm-forming activities of JM109, ATCC 43890, and M17, we applied standard microtiter plate testing. Biofilm formation by bacteria growing in 96-well plates was monitored with crystal violet staining. In striking contrast to the conditions of magnetic levitation, JM109 and M17 strains formed surface-attached biofilms, whereas the strain ATCC 43890 exhibited no noticeable biofilm formation even after 48 h of growth (Fig. 8). To obtain more evidence of matrix production, we applied the same approach used for nonattached levitating aggregates and grew bacteria on the LB agar supplemented with Congo red. Consistent with microtiter plate testing data but in contrast to the data obtained under magnetic levitation conditions, the JM109 and M17 strains grown on agar exhibited Congo red dye binding and produced a light red culture, whereas ATCC 43890 exhibited less efficient Congo red binding. Therefore, the obtained results demonstrated that the relative abilities to form biofilms and levels of matrix component production are different for bacteria grown in nonattached aggregates and surface-attached cultures. These data suggested that the formation of surface-attached biofilms and nonattached aggregates occurs via at least partly different mechanisms.

FIG 8.

Biofilm formation by three E. coli strains. (A) E. coli strains ATCC 43890, M-17, and JM109 were tested in microtiter plate assays. Biofilm biomass was determined using the crystal violet assay. ATCC 43890 did not form representative biofilms that could be detected by this method. (B) Bacterial motility was determined 16 h after seeding bacteria into the semiliquid agar. Data represent the mean and SD from three independent experiments. (C) Congo red curli assay. E. coli were grown on the agar supplemented with Congo red for 48 h. M-17 and JM109 but not ATCC 43890 were stained red, which is characteristic of curli-producing bacteria.

DISCUSSION

Here, we described a model that allowed in vitro studies of nonattached bacterial aggregates. The described technique is based on bacterial growth under magnetic levitation conditions where the magnetic force counterbalances the gravitational force. The best-known example of magnetic levitation is an experiment by Geim and Berry, who made a frog levitate in a magnetic field (67). This experiment confirmed theoretical conclusions that the magnetic force that pushes living objects, which are all diamagnetic, out along the gradient of the magnetic field to the point of the lowest magnetic field might be strong enough to balance the gravity.

In their experiment, Geim and Berry used a high magnetic field of 16.5 T and a magnetic field gradient of approximately 8 T m⁻¹. To obtain a levitation effect under a lower magnetic field, Parfenov et al. suggested a system that employed a gadobutrol-complemented paramagnetic medium, and these properties are opposite to the diamagnetic properties of cells (54). The system used two permanent magnets with the maximal field near the magnets of 1.3 T and the gradient of the magnetic field of 2.2 T cm⁻¹. Thus, the lowest magnetic field is approximately a few decimal values of Tesla, which is still greater than Earth’s magnetic field, which is approximately 0.00005 T, but less than the field used in magnetic resonance imaging (MRI) (58, 59). The described system was applied to obtain viable scaffold-free microtissue constructs using tissue spheroids (54). Cell viability in the microtissue constructs suggested the low toxicity of the suggested conditions for mammalian cells.

In the present work, this system was applied to assess bacterial behavior under low-shear conditions of levitation within the volume of the medium. Our results demonstrated that under magnetic levitation conditions, stationary-phase bacteria were clustered within the restricted volume of the medium. The size and the position of the bacterial cluster were dependent on gadobutrol concentrations, which is consistent with the principles of magnetic levitation described above given that a decrease in gadobutrol concentrations decreased the paramagnetic properties of the medium. After the magnetic field was removed, bacteria were spread over the entire space of the flask. Generally, such a behavior resembled the behavior of the overnight culture left on the table, with the exception that bacteria left on the table do not levitate by partly precipitating on the bottom but spread over the volume after shaking.

Bacteria growing under magnetic levitation conditions behave differently. Bacteria grew as an aggregate that subsequently consisted of smaller mini- and microaggregates formed by bacteria embedded into the polymeric matrix. Formation of aggregates was not just a consequence of magnetic forces that directed bacteria to the point of the lowest magnetic field, because free-floating bacteria outside the aggregate bacteria accounted for approximately 2 to 10% of the total population. Gadobutrol itself was not a causative agent of bacterial aggregation, because growth in the paramagnetic LB medium outside the magnetic bioprinter did not result in microaggregate appearance. Rather, aggregate formation was an inevitable process of bacterial growth in low-shear conditions with no surface for attachment. Thus, magnetic levitation conditions mimic an infection-related environment that occurs in the restricted space of alveoli, mucosal tissues, and cytoplasm (4, 16, 34, 35, 39, 41, 42, 44).

CLSM and SEM revealed that aggregates obtained under magnetic levitation conditions morphologically resembled surface-attached biofilms. Another feature that made the levitating aggregates similar to biofilms was the strong requirement for bacterial growth for aggregate formation. Finally, bacteria in levitating aggregates demonstrated enhanced antibiotic resistance, which is a hallmark of bacterial biofilms. These three biofilm-like features were reported in nonattached aggregates in previous studies (32, 33, 35–42).

Taking advantage of easy manipulation using the model, we revealed and quantified differences between E. coli strains to form nonattached aggregates and biofilms. The strain JM109, which is widely used for biofilm experiments, exhibited poor nonattached aggregate formation. The strain ATCC 43890, which was chosen for its strong nonattached aggregates, failed to form biofilms on plastic and agar surfaces. Congo red matrix labeling suggested that the lack of a matrix is the major reason for both poor biofilm formation and poor nonattached aggregate formation. The mechanism underlying differences in matrix production under different growth conditions is a subject of further research.

Taken together, the results present a magnetic levitation system for in vitro studies of nonattached bacterial aggregates. The system allowed real-time monitoring of aggregate growth and quantification of strain ability to form aggregates. The data obtained with the system demonstrated that a good biofilm formation ability does not necessarily mean a good nonattached aggregate formation ability, and vice versa. A strain that exhibits a good ability to form nonattached aggregates might be ineffective in biofilm formation. Thus, the obtained results support previously published opinions on limitations of biofilm models to study the behavior of human pathogens (14, 43). Moreover, the present data suggest a means to overcome these obstacles. Taking into account the growing understanding of the role that nonattached bacterial aggregates play in the pathogenesis of chronic diseases, the developed system might be useful for modeling bacterial behavior in chronic infections.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Three strains were used in these studies, the Shiga-like toxin 1-producing strain ATCC 43890 (O157:H7), the laboratory strain JM109 (endA1 recA1 gyrA96 thi hsdR17 [r_k⁻ m_k⁺] relA1 supE44 Δ[lac-proAB] [F′ traD36 proAB lacI^qZΔM15]), and the probiotic strain M17 (serotype O2:H6 [68]). Bacteria were routinely cultivated in LB nutritional medium (broth or agar) at 37°C. To achieve magnetic levitation, bacteria were grown in the paramagnetic LB medium, i.e., LB broth supplemented with the paramagnetic substance gadobutrol (see below).

Magnetic bioprinter.

The scheme of the magnetic bioprinter is shown in Fig. 1A. The main elements of the bioprinter include a magnetic unit (Fig. 1A, position 2), a portal for installing a cuvette (Fig. 1A, position 4), and a case (Fig. 1A, position 2). The magnetic bioprinter creates a nonuniform magnetic field in the working area where the cuvette is placed (Fig. 1E). The structure of the magnetic field is formed due to a special construct consisting of two magnetic rings combined by poles of the same polarity (Fig. 1C, position 1). To align the magnetic bioprinter, it was fixed with a cover (Fig. 1C, position 3), which was subsequently connected to a box with screws. To monitor the process in the course of the experiments, a viewing window was arranged for direct visualization of behavioral changes of the bacterial population (Fig. 1A, position 3). Briefly, 5-ml sterile syringes were used as cuvettes. Two independently created magnetic bioprinters were used randomly in the experiments (Fig. S3). The magnetic field parameters were similar for the two devices.

Paramagnetic nutritional medium.

To create a paramagnetic nutritional medium, LB broth was aseptically supplemented with 20% (vol/vol) (or 10% in some experiments) Gadovist (Bayer). The active substance of Gadovist is 1 M gadobutrol ([10-[2,3-dihydroxi-1-(hydroxymethil)propil]-1,4,7,10-tetraazacyclododexan-1,4,7-triaceto(3-)-N¹, N⁴, N⁷, N¹⁰,O¹, O⁴, O⁷]gadolinium). Gadobutrol is a paramagnetic substance used for diagnostic purposes in magnetic resonance imaging (MRI) studies. Briefly, 10 or 20% Gadovist corresponds to 0.1 M or 0.2 M gadobutrol, respectively.

Bacterial growth in conditions of magnetic levitation.

In the standard experiment, an isolated E. coli colony was used to seed LB broth. The overnight culture grown at 37°C with shaking was diluted 1:100 with the sterile paramagnetic nutritional medium. The sterile 5-ml syringe was filled with 2 ml of bacterial suspension, and the extra air was removed. The tip of the filled syringe was heated and fused. The syringe was placed into the working area of the magnetic bioprinter. The bioprinter itself was placed into the 37°C incubator.

Quantification of nonattached aggregates.

Three methods were used to quantify nonattached aggregates formed by bacteria grown in the magnetic bioprinter in the paramagnetic nutritive medium. First, the macroscopic geometric parameters of the levitating aggregates were determined using photos obtained through the window in the magnetic bioprinter. Second, the medium was removed using the syringe, and the aggregate was suspended in PBS and vigorously shaken for 30 s to separate bacterial cells. Then, the decimal dilution of the suspension was plated on the LB agar. Colonies that grew were counted 24 h later. Separation of bacterial cells was monitored by Gram staining. The third method was based on Congo red staining of amyloid proteins and cellulose and allowed quantification of the bacterium-produced matrix. Briefly, the Congo red (CR) dry powder was added to the freshly prepared LB broth up to 100 μg ml⁻¹. Then, the solution was sterilized at 121°C for 20 min, cooled down to 45°C, and stored at this temperature. Immediately before use, the LB/CR broth was supplemented with 20% gadobutrol to get paramagnetic LB/CR medium. Overnight E. coli culture was diluted 1:100 of the paramagnetic LB/CR red medium, placed into the magnetic bioprinter, and incubated at 37°C. The aggregate and the surrounding medium were placed into the 15-ml centrifuge tube, and bacterial cells were concentrated by centrifugation at 4,200 rpm for 10 min. Then, bacteria were washed with 1.5 ml PBS and concentrated by centrifugation as described above. Briefly, 1.5 ml of 50% ethanol was added to the pellet, and the bacterial suspension was rigorously mixed by vortexing for 15 s. Then, bacteria were removed by centrifugation. Briefly, 300 μg of the supernatant was transferred into the wells of a 96-well microplate, and the concentration of the dye was detected by optical density of the solution measured at the wavelength of 450 nm with the iMax reader (Bio-Rad).

Visualization of the aggregate structure.

To fix bacteria grown in the magnetic bioprinter, glutaraldehyde was added directly to the sample up to a concentration of 2.5%. After 2 h of incubation at room temperature, the sample was transferred to the centrifuged tube, and aggregates were sedimented by centrifuging at 42,000 rpm and washed with PBS thrice for 10 min. The structure of the aggregates was analyzed using confocal laser scanning microscopy and scanning electron microscopy (CLSM and SEM, respectively). Cell nucleoids were visualized with SYBR green I. Biofilm matrix components were visualized with FilmTracer SYPRO Ruby biofilm matrix stain. A Zeiss LSM 510Meta confocal microscope (Carl Zeiss, Germany) and a Plan-Apochromat 63/1.4 Oil DIC objective were used to acquire the images. The images were processed using Zeiss LSM 510Meta software version 3.2. For SEM aggregate analysis, samples were prepared according to the standard procedure used for biofilms (69) and sputter-coated with 20‐nm‐thick platinum. Camscan S2 (Cambridge Instruments, UK) in the secondary electron imaging (SEI) mode with a 10‐nm optical resolution and an operating voltage of 20 kV was used to acquire images. The images were captured using MicroCapture software (SMA, Russia).

Antibiotic resistance assay.

The MICs of gentamicin (Fluka, USA) were determined using broth dilutions. MICs were 10 μg ml⁻¹, 10 μg ml⁻¹, and 140 μg ml⁻¹ for ATCC 43890, JM109, and M17, respectively. To determine changes in antibiotic resistance attributed to the formation of nonattached aggregates, gentamicin concentrations that were 2-fold MIC were added to the aggregate. After 24 h, viable cells were counted by plating on the LB agar as described above.

Microtiter plate assay for biofilm formation.

To test the potential of bacteria to form surface-associated biofilms, a standard microtiter plate assay was used (70). An overnight E. coli culture was diluted 1:100 in fresh LB broth and incubated in a 96-well microtiter plate for 48 h. Then, wells were washed with PBS and stained with 0.1% crystal violet solution for 10 min. To assess biofilm biomass, the dye was solubilized by adding 95% ethanol to each stained well. Then, the solution was transferred into a flat-bottom 96-well plate for optical density measurements at a wavelength of 500 nm using the iMark spectrophotometer (Bio-Rad).

Congo red surface-attached biofilm assay.

Congo red (CR) binding was used to reveal amyloid proteins associated with biofilms (65). The CR was added to the LB agar up to a final concentration of 25 μg ml⁻¹. Bacteria were plated using a bacteriological loop and incubated for 24 h at 37°C. Binding was revealed as red labeling of bacterial culture due to CR binding to amyloid curli structures.

Statistics.

All experiments were replicated 3 to 5 times. The mean and standard deviation (SD) values were calculated from the entire data set. Student’s t test results with P < 0.05 were considered statistically significant.

Data availability.

All data that support the findings of this study are either included in this published article or available from the corresponding author upon request.