Protozoa in the environment have been implicated as both potential hosts harboring pathogens (Barker and Brown, 1994) and as agents enhancing pathogen survival and pathogenicity (Rasmussen et al., 2005; Bichai et al., 2008). Presence of the shiga toxin-encoding prophage in Escherichia coli O157:H7 (EcO157) is reported to enhance their survival in the food vacuoles of grazing Tetrahymena pyriformis (Steinberg and Levin, 2007). The passage of Salmonella enterica through and excretion from a soilborne Tetrahymena species is reported to convey increased survival of the organism (Brandl et al., 2005). However, protozoan predation was linked to decreases in EcO157 populations in dairy wastewater (Ravva et al., 2010, 2013) but only three ciliate protozoa were isolated in pure culture (Ravva et al., 2010). For these reasons and difficulties in culturing environmental protozoa, it is important to be able to identify protists in the environment which are actively ingesting bacteria and providing a safe-haven for environmental persistence and transport of EcO157 and other human enteric pathogens.

The study of bacterivory by protozoans by flow cytometry presents several challenges, not the least of which is the dynamic range demand on the instrument. In order to discern bacteria from protozoa, they must bear some separately identifiable characteristic. Size is an obvious choice and the different populations can be observed according to their different locations on a side scatter vs. forward scatter plot. Logarithmic scaling is required to overcome the wide variation in size between the bacteria and protozoa. Creative adjustment of signal threshold and detector voltage levels is required when the ratio of protozoa to bacteria is low (Rifa et al., 2002).

However, this gives us no information about protozoa that have ingested bacteria. Generally, this means that the bacteria are labeled with some fluorescent or otherwise identifiable compound. Examples have included chemical staining of the prey bacteria (Gonzalez, 1999; First et al., 2012), in vivo expression of fluorescent proteins (Fu et al., 2003) and even bioluminescence (Nelson et al., 2003).

Nonetheless, it is commonly observed that a variety of living cells also exhibit a degree of fluorescence, i.e., auto fluorescence, in the absence of any added fluorescent label. Protozoa typically being much larger than bacteria, they can be weakly auto fluorescent and yet on a cellular basis produce a signal of the same magnitude or greater than a single fluorescently labeled bacterium (First et al., 2012).

These problems are further compounded when studying organisms in environmental samples where the presence of other matter, living and otherwise, makes the identification and sorting of target populations even more difficult. In such cases, the use of nucleic acid staining has proven useful in locating the target organisms in the detritus (Lindström et al., 2002).

In this paper we examine some of these issues using a ciliate protozoan Tetrahymena species and pathogenic EcO157 transformed to express green or red fluorescent proteins (GFP or RFP).

We compared both GFP and RFP EcO157 for the grazing studies and found that there was no significant advantage of one over the other. Both strains gave good signal strength but the Tetrahymena was observed to autofluoresce with approximately equal intensity to the average single bacterial fluorescence signal in either case (data not shown). Because of this autofluorescence, the uptake of bacteria is preferably several-fold in order to significantly resolve the bacterivorous protists from the non-feeding population.

The plot in Figure 1 shows a typical time course of fluorescence uptake by Tetrahymena feeding on GFP-EcO157 with an initial concentration of 10⁶ mL⁻¹. Neither heat nor formalin fixation had an adverse effect on GFP fluorescence of the individual E. coli bacteria. Therefore, either method could be used to inactivate the pathogenic bacteria for safety concern when sorting but we chose to use formalin fixation subsequent to feeding so that the bacteria could be fed to Tetrahymena live. In any case, the plot shows a rapid uptake of EcO157 whether live or heat killed, the latter yielding somewhat more peak fluorescence internalization.

As Tetrahymena feeds on EcO157 cells, there is a broadening of the green fluorescence of the population and a shift to higher average fluorescence (Figure 2). Because neither Tetrahymena nor the GFP-EcO157 cells have a narrow fluorescence distribution (fluorescence due to autofluorescence in the former case and GFP in the latter) we do not see discrete increases in fluorescence, i.e. it is not possible from the histograms to discern separate populations according to the numbers of bacteria ingested per cell.

The overall fluorescence of a protozoan cell at any given time is determined by the rate at which it is ingesting fluorescent bacteria, the rate at which it is excreting or digesting fluorescent bacteria, the rate at which the GFP is inactivated inside the Tetrahymena as well as the cell's autofluorescence. The ingestion rate is expected to be a function of bacterial concentration and at low enough concentration, the rates of digestion and/or excretion may predominate such that significant fluorescence accumulation is not possible.

Indeed, we found that the initial concentration of prey is a vital parameter in determining the maximum uptake of fluorescence. This is clearly shown in Figures 3, 4. At the lower initial concentrations of 10⁴ and 10⁵ bacteria mL⁻¹, the feeding Tetrahymena did not develop enough fluorescence to be clearly discerned from the non-feeding population. On the other hand, as Figure 4 shows, there were very few cells which had no increased fluorescence when the initial bacterial concentrations were 10⁷ and 10⁸ bacteria mL⁻¹. The mean fluorescence of the individual E. coli and unfed Tetrahymena in this experiment were 18 and 16, respectively, on the scale shown.

We sorted Tetrahymena with high and low levels of green fluorescence. Figure 5 shows confocal images of example cells sorted from these regions verifying the presence and absence of ingested GFP-EcO157 in the two sorted groups. In this example, the main confounding parameters were the high concentration of background bacteria which were 1000-fold more prevalent (10⁷ EcO157 cells: 10⁴ protozoa mL⁻¹) than the Tetrahymena beside the approximately equal fluorescence intensity of the empty Tetrahymena and the single GFP-EcO157 cell. However, even when we added additional debris the feeding Tetrahymena could be identified from the background.

The feeding of fluorescent protein labeled prey bacteria is a useful tool to identify and sort bacterivorous protozoa, at least in the case of the ciliate Tetrahymena. The ultimate level of fluorescence incorporation is determined by the rates of uptake, excretion and loss due to digestion. It is reported that GFP is unstable and loses fluorescence in the acidic food vacuoles of protozoa even when the bacteria are not completely digested (Parry et al., 2001). We found the fluorescence to be sufficiently persistent in fixed cells to be analyzed and sorted by FACS as well as to be visualized under confocal microscopy. On the other hand, if GFP was persistent in undigested, intact bacteria excreted by Tetrahymena (Brandl et al., 2005; Steinberg and Levin, 2007), we should have been able to identify fluorescent vesicles but this was not the case during a 4 h incubation in this study. However, we did observe vesicles filled with GFP-EcO157 (the same strain) expelled from environmental protozoa 2 days after ingestion (Ravva et al., 2010) and internalized GFP-EcO157 cells were detected in protozoa even after 14 days.

One of the hurdles in utilizing this technique is the incorporation of sufficient fluorescence in the protozoa to obtain a separately identifiable population from within environmental samples. This can be overcome by using a sufficient concentration of labeled prey that gives signal significantly higher than the autofluorescence of the target protozoa and other biological and non-biological components. In the environment, the concentration of host and prey would almost certainly be lower than the levels used here. But there seems to be a concentration of prey below which bacterivory is disfavored or at least below which the ingestion rate is too slow relative to excretion and digestion to allow sufficient accumulation of fluorescence to discern. This observation is consistent with the report of Steinberg and Levin that Tetrahymena grazing on EcO157 ceases below concentrations of about 10⁵ bacteria mL⁻¹ (Steinberg and Levin, 2007) which they attribute to minimum consumption requirement for growth (Watson et al., 1981).

Ingestion rates have been measured by using uniformly fluorescent microspheres (Lavin et al., 1990). Because of the high uniformity of such particles, populations with discrete numbers of ingested particles can clearly be discerned on fluorescence histograms of the protozoa. Although it is not possible with GFP-EcO157 which are not so uniform in size and fluorescence, this method may be used to determine within 5–45 min if environmental protozoa are responsible for significant decreases (Ravva et al., 2010, 2013) of enteric pathogenic bacteria in agricultural environments. Since, isolation and characterization of protozoa in the environment is extremely difficult for use in direct tests to measure the uptake of pathogenic bacteria, flow cytometry appears to be an ideal method in sorting and culturing protozoa based on their relative uptake of fluorescence-labeled organisms. We used flow cytometry earlier to characterize bacteria from urban aerosols (Hernlem and Ravva, 2007). Feasibility studies with environmental samples are needed to determine if protozoa preferentially harbor and transport enteric pathogens from point sources (dairies and feed lots) to fruit and vegetable crops grown in proximity.