The bacterial species X. retroflexus (JQ890537), S. rhizophila (JQ890538), M. oxydans (JQ890539), and P. amylolyticus (JQ890540) stored in the culture collection of the Section of Microbiology, University of Copenhagen, were subcultured from frozen glycerol stocks onto TSA plates (Tryptic Soy Agar, Sigma–Aldrich, Germany). The plates were incubated at 24°C for 48 h. Single colonies were inoculated into 5 ml TSB (Tryptic Soy Broth, Sigma–Aldrich, Germany) media and incubated with shaking at 180 rpm for 24 h at 24°C when required. These strains were used as bacterial prey for the protozoan predator.

A T. pyriformis (Tp) culture (Culture Collection of Algae and Protozoa, CCAP nr 1630/w1) was provided by Department of Veterinary Public Health and Food Safety, Ghent University, Belgium. Axenic cultures of this protozoan were maintained in 25 cm² culture flasks with 20 ml PPY medium [proteose peptone yeast extract; 20 g Proteose peptone (Merck KgaAm Germany), 2.5 g yeast extract (Merck KgaAm Germany) in 1 L H₂O; autoclaved]. Weekly maintenance of the ciliate cultures at 24°C was done by aseptically transferring 5 ml of the culture into 15 ml fresh PPY medium incubated. For biofilm grazing experiments, T. pyriformis cells in exponential phase (after 48 h at 24°C) were washed twice in PAS (Page’s amoeba saline) solution followed by centrifuging at 850 g after which the cells were re-suspended into 10 ml TSB media.

Biofilm cultivation experiments were performed in 96-well cell culture plates (cat. no. 655180, Greiner Bio One, Germany). The four selected strains were screened for biofilm formation as single species (monospecies) and in three/four-species combination (multispecies) as described (Ren et al., 2015) both in the presence and absence of protozoa. Briefly, bacterial cell cultures in exponential growth phase (OD₆₀₀ between 0.3 – 0.6) were selected and adjusted to a start OD₆₀₀ of 0.15 in TSB media for all cultures. For monospecies biofilms, aliquots of 150 μl of cell culture and for three- and four-species biofilms, respectively, 50 or 37.5 μl of each bacterial strain were added into the wells so that the final inocula were 150 μl in all the settings. To the wells that were to be grazed, an volume of 1.5 μl containing ∼approximately 1000 cells T. pyriformis cells in TSB media were added. The plates were incubated at 24°C for 12, 24, and 96 h. Wells containing only 150 μl TSB media and TSB media with T. pyriformis cells served as blank/control. Three wells each time served as one technical replicate and this was repeated at five different times.

Biofilm formation was assayed and quantified using the traditional crystal violet (CV) method as previously described (Ren et al., 2014). The biofilm attached to the wells was then washed twice gently with 160 μl 1X PBS (phosphate buffer saline) solution and stained with 180 μl 1% (w/v) CV solution. After 20 min of staining, the CV solution was removed by pipette, and the stained biofilm was gently washed five times with 200 μl PBS solution. The remaining CV dye retained by the biofilm was de-stained into 200 μl 96% ethanol for 30 min. Biofilm formation was then quantified by measuring the absorbance of de-stained CV at 590 nm using EL340 BioKinetics reader (BioTek Instruments, United States) and expressed as biofilm forming index (BFI) according to the equation BFI = (AB-CW)/G (Niu and Gilbert, 2004) where, AB: OD₅₉₀ of attached microorganisms, CW: OD₅₉₀ control wells and G: OD₆₀₀ of cells in planktonic fraction. Biodiversity (BD) effect was calculated as the difference between the observed biofilm yield (biofilm of mixed cultures) and the expected yield (average of the monoculture yields) (Loreau and Hector, 2001; Vanelslander et al., 2009). Biofilm fold (F_d), i.e., the observed increase in biofilm formation due to grazing is the ratio between OD₅₉₀ of grazed three-species biofilm and OD₅₉₀ of non-grazed three-species biofilm.

Quantification of the biofilm and planktonic fractions was performed by plating. 100 μl of the planktonic fraction from the wells after 24 and 96 h incubation was suspended in 900 μl 1X PBS solution. Once the planktonic fractions were removed, the wells with attached biofilm were gently washed twice with 160 μl 1X PBS solution. The wells were then filled with 200 μl 1X PBS and mixed thoroughly by pipetting. Serial dilutions in 900 μl 1X PBS were performed and 100 μl of the dilutions were plated onto TSA plates by spread plating after which the plates were allowed to dry completely at room temperature. Drying restricts the movement of T. pyriformis on plates. The plates were then incubated for 48 h at 24°C. Single colonies formed after incubation were counted and the results were calculated in CFU (colony forming units). Grazing fold, i.e., the percentage reduction in planktonic fraction due to grazing was expressed by 100 × [CFU_(culture) - CFU_(culture+Tp)/CFU_(culture)]. Changes in cell counts from biofilm fraction were expressed by log [CFU_(culture+Tp)/CFU_(culture)].

We determined the ciliate numbers of T. pyriformis grown on the four bacteria separately (monospecies) and as a mixture (four-species) for up to 96 h at regular intervals in microtiter plates. The protozoa cells were counted using a Sedgewick-Rafter chamber and an inverted microscope (40× magnification) as described previously (Gittleson and Ganapathy, 2011) with minor modifications. The wells containing the suspension of bacteria and protozoa were homogenized by pipetting and 150 μl of the cell suspension was fixed in 1% (w/v) Lugol’s iodine solution to a final volume of 1.2 ml in dH₂O. The contents were then immediately transferred to the counting chamber and the cells were allowed to settle for few minutes. The change in protozoa cell numbers over time was expressed using ΔN = log₁₀ (N_t -N₀)/t. To visualize the changes in protozoa numbers over time in co-culture with bacteria, 50 μl spots of the fixed suspension were made on glass slides and micrographs were taken at different time points using Zeiss Axioplan II, Carl Zeiss with a 10× objective.

To visualize the effects of grazing by the protozoan and the internalization of bacteria within the food vacuoles of T. pyriformis, FISH was performed with 16S rRNA gene probes targeting the specific bacteria (Liu et al., 2017). 50 μl spots of co-culture suspensions (bacteria and protozoa) after 24 h were collected after thorough pipetting to homogenize the suspension. The collected cells were then left to air dry on a glass slide. The above step was repeated five times (5 μl × 50 μl) with the aim to collect more cells. The attached cells were coated with 0.5% (w/v) agarose by immersing the slides into a tube containing 45 ml molten agarose and fixed using 4% PFA (paraformaldehyde) at 4°C. Samples were dehydrated and the hybridization protocol was performed according to (Amann, 1995; Daims, 2009) with 30% formamide concentration. After hybridization, the slides were washed in cold water and dried at room temperature. The slides were stored in the dark and visualized under confocal microscopy (Point-scanning confocal and multiphoton microscope SP5-X MP, Leica Microsystems). Images were processed using Leica Application Suite X.

The biofilm formation assay was conducted both in the presence and absence of T. pyriformis in 96-well microtiter plates as described above. After 24 h, the planktonic fractions were collected in Eppendorf tubes and the biofilm fraction was rinsed twice with weak phosphate buffer to remove loosely attached cells. Three replicate wells were prepared for each treatment. The cell numbers of the four strains in multispecies planktonic and biofilm fractions with and without protozoa were quantified by SYBR Green qPCR using standard curves generated by serial 10-fold dilutions of plasmid DNA using the species specific primers and thermal profile setup previously reported (Ren et al., 2014). All samples were run in triplicate and a no template control was included in each run. Bacterial DNA was extracted using FastDNA^TM SPIN Kit for soil (MP Biomedicals, Germany) according to manufacturer’s instruction.

Monocultures and four-species mixed cultures of X. retroflexus, S. rhizophila, M. oxydans, and P. amylolyticus were tested for biofilm formation in the absence and presence of protozoa (Figure 1). T. pyriformis grazing on monospecies cultures of X. retroflexus and S. rhizophila resulted in significantly enhanced biofilm formation (paired t-test, P < 0.05) whereas M. oxydans and P. amylolyticus monocultures did not form biofilms neither in the presence nor in the absence of T. pyriformis. Biofilm formation was enhanced in the four-species mixtures, and was even more strongly induced in these mixtures in the presence of grazing for up to 96 h (n = 5, paired t-test, P < 0.05), suggesting a strong biodiversity effect (Supplementary Figure S1). Moreover, biofilm formation in the mixtures was higher than in the best performing monoculture both in the absence and presence of grazing.

Planktonic fractions of three of the four bacterial species were effectively grazed upon in monoculture, but less so in the four-species co-culture. P. amylolyticus was the most intensively grazed species at 24 h, whereas after 96 h S. rhizophila monocultures were the most highly grazed followed by P. amylolyticus and X. retroflexus monocultures. Among all the strains, M. oxydans was the least preferred prey, and S. rhizophila and P. amylolyticus were the most favored prey (Figure 2). These grazing experiments verified the ability of T. pyriformis to feed on planktonic bacteria. In the four-species mixed cultures, overall grazing by T. pyriformis on the planktonic community was reduced compared to the monospecies cultures observed by the low grazing fold values at 24 and 96 h (Figure 2).

In the biofilm fraction, cell numbers of X. retroflexus and S. rhizophila increased at 24 and 96 h in the grazed relative to the non-grazed monocultures whereas the cell numbers of M. oxydans and P. amylolyticus decreased with grazing compared to the non-grazed monocultures (Figure 3). This underscores the inability of M. oxydans and P. amylolyticus to form a biofilm in monoculture. In the four-species culture, total cell numbers increased both at 24 and 96 h compared to the non-grazed biofilm (Figure 3 and Supplementary Figure S2).

The growth of T. pyriformis cells on all bacterial isolates cultured as both mono and mixed planktonic cultures was followed over time (Figure 4). The change in cell numbers over time demonstrated that S. rhizophila and P. amylolyticus were suitable prey for the protozoa (Figures 4B,D) and that TSB media can support the axenic growth of protozoa (Figure 4F). Growth on M. oxydans was not pronounced (Figure 4C); while X. retroflexus monocultures had a negative impact on the growth of the protozoa at 96 h (Figure 4A). Our results thus indicate that T. pyriformis may prefer to graze on S. rhizophila and P. amylolyticus. The numbers of protozoa grazing on the four-species mixed cultures represent a smoother curve over time indicating that the protozoa can adapt to an available prey in multispecies bacterial environments (Figure 4E).

To visualize the change in protozoan numbers over time, micrographs showing T. pyriformis cells raised on both mono and multispecies bacterial cultures are shown (Supplementary Figure S3). The protozoan population raised on the four-species mixtures remained viable for up to 96 h. However, in monospecies cultures; it can be seen that the protozoan cell numbers increased from 24 h and reached a maximum at 96 h when co-cultured with S. rhizophila and P. amylolyticus whereas the protozoa population declined from 24 to 96 h in co-culture with X. retroflexus and M. oxydans.

To visualize grazing on monocultures and mixed cultures, a 16S rRNA gene based FISH was performed, similar to a previous study (Jezbera et al., 2005), after 24 h of grazing and samples were visualized by laser scanning confocal microscopy. It was confirmed that T. pyriformis can consume the bacteria in all tested monospecies settings, however, at seemingly different rates as indicated by the number of food vacuoles formed within the ciliates (Figure 5). In co-cultures of T. pyriformis with X. retroflexus or S. rhizophila monocultures, the bacteria were abundantly present within the food vacuoles of T. pyriformis cells (Figures 5A,B) showing that these bacterial strains are readily consumed. P. amylolyticus cells were also found to be localized within the food vacuoles of grazers but not as abundantly as compared to X. retroflexus and S. rhizophila (Figure 5D). Most protozoan cells appeared to form cysts when co-cultured with M. oxydans (Figure 5C), but some bacterial cells were found to be internalized within T. pyriformis indicating that the protozoa were able to consume M. oxydans cells. In the case of grazing on four-species mixed cultures (Figures 6A–H), most food vacuoles were dominated by X. retroflexus indicating that at 24 h most protozoan cells prefer to graze on X. retroflexus. This was in accordance with the fact that this bacterium previously was shown to dominate the 24 h mixed biofilm population (Ren et al., 2014) and thus could be readily available for the grazers.

From the above results, biofilm formation in the presence of T. pyriformis was further assessed to better understand the dynamics. To this end, either the least preferred prey M. oxydans or the best biofilm producer X. retroflexus were excluded three-species consortia (Figure 7). Biofilm formation (biofilm-fold F_d) was enhanced when X. retroflexus remained in the consortium together with S. rhizophila and P. amylolyticus, indicating that the interaction between these three members is vital for biofilm stability. However, in the absence of X. retroflexus and in the presence of M. oxydans, the consortium was effectively grazed, although there seemed for this consortium to be a gradual adaptation to predation (as evidenced by increased biofilm formation) over time.

In order to determine the cell numbers of the individual species within the multispecies consortium, 16S rRNA gene based q-PCR quantification was applied according to a previously developed protocol (Ren et al., 2014). The results showed that in the multispecies biofilm fraction, cell numbers of X. retroflexus, S. rhizophila, and P. amylolyticus increased in the presence of grazers compared to the control biofilms that were not grazed. The ∼2.5-fold increase in cell numbers of X. retroflexus and P. amylolyticus and 1.7-fold increase in S. rhizophila cell numbers suggest that synergistic interactions between these species were enhanced in the presence of grazing, resulting in increased cell numbers in the biofilm. The cell numbers of M. oxydans in the biofilm remained unaffected either in the presence or absence of grazers (Figure 8A).

In the planktonic fraction without grazing, a similar trend in cell numbers compared to the non-grazed biofilm was seen with X. retroflexus, P. amylolyticus, and S. rhizophila being the dominant species (Figure 8B). However, the planktonic cell numbers of these species decreased in the presence of T. pyriformis indicating an effect of grazing on these planktonic fractions. In contrast, the cell numbers of M. oxydans increased, which possibly can be a result of grazing preference of the ciliate in the mixed communities and/or higher nutrient or space availability for M. oxydans cells as the other members of the consortia were grazed upon.