Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Microbiology (2016), 162, 610–621 DOI 10.1099/mic.0.000259 The O-antigen mediates differential survival of Salmonella against communities of natural predators Aletheia Atzinger, Kristen Butela3 and Jeffrey G. Lawrence Department of Biological Sciences, University of Pittsburgh, Pittsburgh, PA 15260, USA Correspondence Jeffrey G. Lawrence [email protected] Antigenically distinct members of bacterial species can be differentially distributed in the environment. Predators known to consume antigenically distinct prey with different efficiencies are also differentially distributed. Here we show that antigenically distinct, but otherwise isogenic and physiologically indistinct, strains of Salmonella enterica show differential survival in natural soil, sediment and intestinal environments, where they would face a community of predators. Decline in overall cell numbers is attenuated by factors that inhibit the action of predators, including heat and antiprotozoal and antihelminthic drugs. Moreover, the fitness of strains facing these predators – calculated by comparing survival with and without treatments attenuating predator activity – varies between environments. These results suggest that relative survival in natural environments is arbitrated by communities of natural predators whose feeding preferences, if not species composition, vary between environments. These data support the hypothesis that survival against natural predators may drive the differential distribution of bacteria among microenvironments. Received 20 November 2015 Accepted 15 February 2016 INTRODUCTION Bacteria are dominant members of every ecosystem, their 1030 individuals outnumbering all other cellular life forms (Whitman et al., 1998). Bacteria are major constituents of soil (Trevors, 2010), sediment (Dale, 1974) and marine and fresh waters (Watson et al., 1977). In addition, they colonize both external and internal surfaces of multicellular eukaryotes – where bacterial cells may be 10-fold more abundant than cells of their host (Human Microbiome Consortium, 2012) – and serve as intracellular symbionts to diverse taxa (Baumann, 2005). Turnover of bacteria in natural environments occurs within days (Luna et al., 2002; Roszak & Colwell, 1987), indicating that mortality rates are higher than expected given the natural rate of senescence (Nyström, 2002, 2007). Major arbiters of population control are bacteriophages and amoebae, which limit bacterial population growth in numerous environments (Jjemba, 2001; Peduzzi & Schiemer, 2004; Pernthaler, 2005; Rønn et al., 2002); additional grazers (such as nematodes) may be present as well (Trevors, 2010). Many predators (including components of immune systems) recognize their prey by virtue of the antigenic 3Present address: Division of Natural and Health Sciences, Seton Hill University, Greensburg, PA, USA. Four supplementary figures Supplementary Material. 610 are available with the online molecules on the surface of bacterial cells. As the most abundant molecule on the cell surface, the O-antigen LPS is a likely candidate used in prey recognition. In enteric bacteria, LPS is synthesized by enzymes encoded by the rfb operon; this region is hypervariable among serovars of Salmonella (Grimont & Weill, 2007). LPS components are potent activators of immune responses (Reeves, 1995), including the induction of predatory white blood cells upon systemic infection of a eukaryotic host by pathogenic bacteria. Therefore, it is not surprising that the structure of O-antigen polysaccharides is hypervariable within many species of pathogenic bacteria (Butela & Lawrence, 2010). Often, frequency-dependent selection mediated by the host immune system is invoked as a mechanism maintaining this diversity (Levin et al., 1988; Yuste et al., 2002). Here, the fitness of a variant allele is inversely proportional to its frequency, preventing both the fixation of common alleles and the loss of rare ones. Yet, diverse O-antigens are also found in species not associated with eukaryotic hosts (Wildschutte et al., 2010). Predators found in natural environments could serve as the selective agents for maintaining antigenic diversity in these cases. In this model, predators in different microenvironments vary in their ability to recognize different antigenic types, thereby conferring high fitness to genotypes they avoid consuming, but only within their resident environment. Consistent with this model, diverse amoebae discriminate among bacterial prey on the basis of their O-antigen polysaccharides (Wildschutte et al., 2004; Downloaded from www.microbiologyresearch.org by 000259 G 2016 The Authors IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 Printed in Great Britain Differential survival against predator communities Wildschutte & Lawrence, 2007). Moreover, predators found in different environments have different feeding preferences (Wildschutte et al., 2004; Wildschutte & Lawrence, 2007). Regardless of their phylogenetic relatedness, predators isolated from the same environment share feeding preferences (Wildschutte & Lawrence, 2007), so that unrelated predators dwelling in the same environment avoid consuming the same antigenic classes of prey. A model was presented whereby these preferences were driven by molecular mimicry (Wildschutte & Lawrence, 2007). Here, predators may recognize environmental epitopes for adhesion; bacterial cells whose antigens resemble these epitopes would not be recognized as potential prey. While these results are limited to only a few cultivable amoebae, they suggest that bacterial strains could outcompete conspecific, but antigenically different, strains in natural environments by virtue of their ability to avoid being eaten by the communities of resident predators, as opposed to any physiological or metabolic adaptations they might carry. This model stands in contrast to those that cite such physiological traits as the basis for differential distribution of bacterial strains among environments (Gordon & FitzGibbon, 1999; Gordon & Cowling, 2003; Walk et al., 2009). We tested this hypothesis by competing physiologically identical, but antigenically distinct bacteria within natural environments. METHODS Bacteria. Except for strains SARB2 and SARB36, all strains were derived from Salmonella enterica serovar Typhimurium strain LT2 (Table 1) and in most cases differed from the WT only in carrying a plasmid expressing a fluorescent protein, and/or having the rfb operon replaced so that a different O-antigen was made. A deletion of the rfb operon – D(rfbP-rfbB)3302 – was created by replacing the region between the gnd and galF genes with the cat chloramphenicol resistance gene, all within a galE2841 : : aatA hisD9953 : : MudJ phs-209 : : Tn10dGn background. This parental strain was transduced to histidine prototrophy using bacteriophage P1 grown on galE6866 : : aph derivatives of antigenically distinct strains of the SARB collection (Salmonella Reference Collection B) (Boyd et al., 1993); these derivatives were created using the method of Butela & Lawrence (2012). Retention of the upstream phs-209 : : Tn10dGn marker limited the introduction of DNA from upstream of the his operon. This strain was back-crossed twice to the parental Drfb-3302 strain, selecting as above, and the galE mutation was repaired via P1 transduction from a Gal+ Drfb strain. The identity of the O-antigen in the isogenic strains was verified by antibody agglutination tests. Plasmids pEGFP, pEYFP, pECFP, pDsRed-Express2 and pKAB9 (Lawrence et al., 2013) were introduced by electroporation and express fluorescent proteins GFP, YFP, CFP, DsRed-Express2 and mKalama1, respectively. Media and antibiotics. Bacteria were propagated at 37 uC in LB with 200 mg ml21 ampicillin to maintain plasmids. MacConkey agar (Difco) was cast with xylose added to 1 % final concentration after sterilization. Kligler iron agar (Difco) was incubated at 37 uC in a Forma anaerobic chamber under 85 % N2, 10 % CO2, 5 % H2. For competitions in natural environments, drugs were added 1 h before and 24 h after the start of the experiment: metronidazole (Sigma-Aldrich) was used at 100 mg ml21, tetramisole HCl (SigmaAldrich) at 500 mM, and ivermectin (Sigma-Aldrich) at 20 ng ml21. http://mic.microbiologyresearch.org Table 1. Strains of bacteria used Strain Genotype LT2 LD901 LD902 Salmonella enterica Typhimurium LT2 rfb SARB1 DrfcC3303 : : cat pEGFP rfb SARB1 DrfcC3303 : : cat pDsRed-Express2 rfb SARB1 DrfcC3303 : : cat pEYFP rfb SARB1 DrfcC3303 : : cat pKAB9 rfb SARB1 DrfcC3303 : : cat pECFP rfb SARB52 pEGFP rfb SARB36 pDsRed-Express2 rfb SARB2 pEYFP rfb SARB20 pKAB9 rfb SARB30 pCFP rfb SARB52 pKAB9 rfb SARB36 pEYFP rfb SARB2 pEGFP rfb SARB20 pECFP rfb SARB30 pDsRed-Express2 rfb SARB59 pDsRed-Express2 Salmonella enterica Agona (Xyl2) Salmonella enterica Newport (Xyl+) rfb SARB3 phs-209 : : Tn10dGn rfb SARB4 rfb SARB20 phs-209 : : Tn10dGn pdu-219 : : Tn10dCm D(orgC-prgH)4301 : : yfp,zeo pEGFP LD903 LD904 LD905 LD906 LD907 LD908 LD909 LD910 LD911 LD912 LD913 LD914 LD915 LD917 SARB2 SARB36 HW31 HW19 LD916 O-serotype* 1,4,[5],12 1,4,12 1,4,12 1,4,12 1,4,12 1,4,12 1,9,12 6,8 3,10 8,20 6,7 1,9,12 6,8 3,10 8,20 6,7 1,3,19 1,4,12 6,8 1,4,12 6,7 8,20 *For strains constructed by moving the rfb operon from SARB strains into strain LT2, O-serotype is designated according to the serotype of the donor strain. Propagation of Caenorhabditis elegans. C. elegans strain N2 Bristol was propagated at 22 uC on nematode growth medium (NGM) plates (Cold Spring Harbor) using Escherichia coli as a feeding strain. For Salmonella competition, strains were grown overnight in LB with 0.01 % glucose, rinsed and mixed in equal proportions prior to spreading 50 ml (*108 c.f.u. total) on NGM plates. An aliquot of several hundred nematodes was introduced via an agar chunk. Plates were incubated for 4 days at 22 uC, allowing the nematodes to consume most of the Salmonella lawn. To retrieve remaining Salmonella, the agar in contact with the introduced piece was discarded and cells were eluted in 0.9 % NaCl, plated on indicator media and grown at 37 uC to discriminate between the two strains. The high temperature of incubation prevented growth of eluted C. elegans. Soil, sediment and fish. Sediment was obtained from Panther Hollow Lake (40.436883 uN, 79.947188 uW) and Homewood Cemetery Pond (40.440249 uN, 79.911991 uW); sites were close to shore but under both leaf litter and *8 feet (2.4 m) of water. Sediment was filtered through *0.5 mm nylon mesh. Soil was obtained from the Fruit Room (40.438949 uN, 79.948101 uW) and Desert Room (40.438535 uN, 79.947509 uW) of the Phipps Conservatory, Pittsburgh, PA; these areas have had the same plant species growing in them for many years, and have experienced no outside animal traffic. Predators were extracted by passing an equal volume of water through the soil and filtering as above. Comet goldfish (Carassius auratus) were obtained from Carolina Biological and propagated for 1 month in a community tank at 30 uC Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 611 A. Atzinger, K. Butela and J. G. Lawrence until 5–7 cm in length. All animal work was performed in accordance with approved Institutional Animal Care and Use Committee protocols. Fish were sacrificed by overdose of tricaine methylsulfonate. Intestinal contents were released into PBS (135 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4 pH 7.4) and filtered through nylon mesh. Intestine contents from four fish of approximately equal cumulative mass (*10 g) were pooled. Heat treatment was performed by submerging 30 ml (soil and sediment extracts) or 7 ml (intestinal contents) aliquots in boiling water for 45 min, cooling on ice and equilibrating to room temperature. Competition in complex environments. Overnight cultures of strains were grown at 37 uC in LB with 200 mg ml21 ampicillin, rinsed and resuspended in 0.9 % NaCl and then added as a mixture to environmental samples to a final concentration of 107 cells ml21. Samples were taken by thoroughly mixing the suspension, removing 2 ml and allowing it to gravity-filter through a 5 mm pore-size mesh. A 0.012 U aliquot of biotin-conjugated rabbit anti-Salmonella antibody (Gene Tex) was added and the sample was incubated at 4 uC for 30 min. Samples were pelleted at 10 000 g for 5 min, washed in PBS, pelleted again, and resuspended in an equal volume of PBS before adding washed beads from the CELLection Biotin Binder kit (Invitrogen) and following the indirect technique for cell pulldown recommended by the manufacturer. Samples were fixed in 1 % paraformaldehyde and stored at 4 uC. Flow cytometry. The samples were run on a Beckman-Coulter CyAn ADP flow cytometer with a 0.3 neutral density filter controlled by Summit 4.3 software (Dako), or on a BD LSRFortessa cell analyser controlled by BD FACSDiva software (BD Biosciences). SYTO 62 nucleic acid dye (Invitrogen) was added at a final concentration of 50 nM 15 min before cytometry. Flow rate was maintained at approximately 1000 events s21. Bacteria-only controls were used to differentiate bacteria from debris and larger cells. Data were analysed using Ferdinand (Lawrence et al., 2013); Z-scoring was used to reduce miscounting events that arose from non-bacterial sources (Fig. S1, available in the online Supplementary Material). Measuring fitness. Conventional methods for calculating fitness determine relative growth rates between different genotypes in the same environment [e.g. as dP/dt5sP(12P)]; here, the proportion of cells (P) is measured over time to calculate the selection coefficient (s). In our case, we wished to measure relative cell fitness of a strain in different environments (with and without predators). To do this, we examined the proportions of each strain at the beginning and end of experimental treatments. The frequency (F) of a strain in its environment was calculated as F i ¼ C i =N ð1Þ where Ci is the count of cell type i and N is the sum of all countable cell types in the experiment, here fluorescently tagged cells. Relative recovery for strain i was defined as R i ¼ F N =F 0 ð2Þ where F0 and FN are the frequencies of the cell type at the beginning and end of the experiment, respectively. Fitness in the face of predators can be calculated as W i ¼ Rp =R A ð3Þ where RP and RA are the relative recoveries in the presence and absence of predators, respectively. 612 Variance of both relative recovery and fitness was calculated by resampling. Estimates of genotype frequency using multiple replicates were used to infer a mean and variance, assuming a Gaussian distribution. These distributions were resampled 1000 times and used to estimate the variance of relative recovery and fitness. RESULTS AND DISCUSSION Salmonella is removed from natural environments by biological agents Many studies have documented the action of predators in the decline of bacterial populations in natural environments (Hahn & Höfle, 2001; Rodrı́guez-Zaragoza, 1994; Rønn et al., 2002; Sørensen et al., 1999). To establish the time frame of predation in our system, we followed the disappearance of Salmonella within a sediment sample from Homewood Cemetery Pond. This strain was discriminated from other bacterial cells by GFP-fluorescence and resistance to chloramphenicol, gentamicin and zeomycin. Prior to strain addition, sediment was either left untreated, treated with heat or treated with a cocktail of three antibiotics targeting eukaryotic predators. Samples were recovered over time and the number of tagged Salmonella cells remaining was determined by plating on selective media; resulting colonies were verified by fluorescent signal from GFP; no colonies were detected from samples that did not receive Salmonella aliquots. As expected, the numbers of introduced cells recovered from untreated sediment decreased over time (Fig. 1a), with fewer than 2 % remaining after 48 h. However, populations showed no significant decline in sediment exposed to either high temperature or a mixture of antiprotozoal and antihelminthic drugs (Fig. 1a). These data suggest that biological agents are responsible for the decline of Salmonella added to the untreated samples. Since drugs targeting eukaryotic predators greatly increased survivorship, we suspect that these organisms are dominant predators in these environments; other environments, such as sea water (Chibani-Chennoufi et al., 2004; Wommack & Colwell, 2000), have dominant bacteriophage predators. Since bacteriophage would be unaffected by antiprotozoal and antihelminthic drugs, they cannot be responsible for the decline in Salmonella numbers in untreated samples. Diverse antigenic types can be recovered by antibody binding For high-throughput enumeration of Salmonella, cells were first recovered from natural environments by binding to biotin-conjugated anti-Salmonella antibody. To determine if all antigenic types used in this study could be recovered efficiently by this antibody, a mixture of five strains – each expressing a different O-antigen and a different fluorescent protein, but otherwise isogenic – was assembled. Ratios of cell types were determined by flow cytometry for samples collected both before and after binding to the antibody Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 Microbiology 162 Differential survival against predator communities (a) cells may reflect differences in the longevity of fluorescent proteins rather than changes in relative abundance of the cell types within the original sample. To test the effect of the fluorescent tags on recovery and enumeration of Salmonella, strains expressing one of five fluorescent proteins (strains LD901–905) were introduced into sediment from Homewood Cemetery Pond; except for the plasmid expressing different fluorescent proteins, cells were otherwise isogenic. After either 0 or 24 h incubation, cells were recovered using Salmonella-specific antibodies and counted via flow cytometry. The same proportions of the five fluorescent strains were found at both time points (Fig. 2a); these results indicate that the half-life of the fluorescent proteins did not significantly differ. Similar results were obtained after 48 h incubation (data not shown), the maximum duration of experiments described here. Fluorescent signal was never found in samples lacking added fluorescent cells. 108 Cell count 107 106 105 104 48 24 Time in sediment (h) 0 Pellet/supernatant frequency (b) 2.0 Nature of the O-antigen affects relative recovery and enumeration 1.5 1.0 0.5 (L D 91 0) 6, 7 (L D 90 9) 8, 20 (L D 91 7) 1, 3, 19 (L D 90 8) 3, 10 1, 9, 12 (L D 90 6) 0 O-antigen serotype (strain) Fig. 1. (a) Survivorship of Salmonella strain LD916 in Homewood Cemetery Pond sediment. Cell count on selective medium is plotted against time. White bars, untreated; black bars, heat-treated; grey bars, drug-treated. (b) Recovery of antigenically distinct strains of S. enterica by antibody pulldown. The recovery of each serotype following antibody binding (pellet) is shown relative to its frequency in a mixed liquid culture (supernatant) for replicates. and recovery using avidin beads. While subtle differences (about twofold) were seen in the efficiency of binding to the antibody among strains expressing different O-antigens (Fig. 1b), these did not prevent robust recovery and enumeration of the different serovars of Salmonella. This antibody recovery protocol was used in all competition experiments. Fluorescent tags do not affect enumeration of cells via flow cytometry High-throughput enumeration of bacterial cells used flow cytometry, where strains were tagged with different fluorescent proteins. Differential enumeration of fluorescent http://mic.microbiologyresearch.org To determine if the identity of the O-antigen affects relative survival of bacteria in natural environments, we added a mixture of fluorescently tagged, antigenically distinct, but otherwise isogenic strains of Salmonella (LD906–910) to a sediment sample collected from the Homewood Cemetery Pond. Cells were recovered by binding to biotin-conjugated rabbit anti-Salmonella antibody after 0 and 24 h incubation and enumerated by flow cytometry. While the nature of the fluorescent tag did not influence relative recovery (Fig. 2a), this was not true for the nature of the O-antigen (Fig. 2b). There were large differences in relative recovery after 24 h (Fig. 2b; P52.26|10210, ANOVA). For example, strains bearing the O-antigens originating from SARB52 (O : 1,9,12) and SARB2 (O : 3,10) both increased in abundance over 24 h, while those bearing O-antigens originating from SARB20 (O : 8,20), SARB30 (O : 6,7) and SARB36 (O : 6,8) decreased. Because the strains were isogenic at all loci except the region containing the O-antigen-encoding rfb operon, the identity of the O-antigen appears to affect relative recovery of Salmonella serovars within natural environments during a time frame when cell numbers decrease (Fig. 1a). Relative survival of antigenically distinct strains is affected by heat-labile factors Differences in relative recovery of antigenically distinct strains after 24 h incubation in natural environments could be driven by differential binding to silicates or interaction with other, non-biotic aspects of the environment. Alternatively, because predation affects overall survival (Fig. 1a), interaction with predators may lead to differences in cell numbers over time. If heat treatment of the environment influences relative recovery of Salmonella, then biotic elements would likely be responsible for differential survival of Salmonella, rather than simply differential recovery. Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 613 A. Atzinger, K. Butela and J. G. Lawrence Relative recovery (f24/f0) (a) (b) 2.0 1.5 1.0 0.5 Fluorescent tag (strain) 0) 91 D (L 7 6, (L D 90 9) 7) 8, 20 90 D (L 6, 8 (L D 90 8) 3, 10 (L D 90 6) 1, 9, 12 Ka l( LD 90 4) C FP (L D 90 5) m (L D 90 2) D SR ed (L D 90 3) YF P G FP (L D 90 1) 0 O-antigen serotype (strain) Fig. 2. Relative recovery of S. enterica from Homewood Cemetery Pond sediment. (a) Relative recovery of strains of S. enterica LT2 (LD901–905) expressing different fluorescent proteins after 24 h incubation in sediment. Relative recovery was calculated as the ratio of the fractions of recovered cells expressing each fluorophore after 24 h to those at 0 h incubation ( f24/f0). (b) Relative recovery of S. enterica LT2 expressing different O-antigens and different fluorescent proteins (LD906–910). Relative recovery of each serotype was adjusted for the differential expression of fluorophores as determined in Fig. 1(a). For clarity, the fraction of serovar 1,9,12 is plotted at 10 % of its measured value. To test this hypothesis, five fluorescently tagged, antigenically distinct but otherwise isogenic strains of Salmonella (LD906–910) were competed within sediment samples collected from Homewood Cemetery Pond. The sediment samples either were left untreated or had been heat-treated before equilibration to 25 uC. The mixture of Salmonella cells was introduced, and cells were recovered by antibody binding after either 0 or 40 h of incubation; cells were enumerated by flow cytometry. As expected, the fractions of each cell type were the same for untreated and heat-treated sediment at the start of the experiment (Fig. 3a; P50.18, ANOVA), indicating that heat treatment did not affect any contribution of differential binding to substances within the sample. However, fractions of each cell type significantly differed between untreated and heat-treated sediment after 40 h (Fig. 3b; P51.1|1029, ANOVA). While strain LD906, bearing the O : 1,9,12 serotype, strongly outcompeted other strains in untreated, predator-bearing samples, this advantage was eliminated when the samples were heat-treated prior to competition, thereby eliminating the predators (Fig. 3b). This suggests that LD906 effectively escaped predation in the untreated samples, while other serotypes were consumed more readily. These differences in abundance over time led to differences in relative recovery between strains (Fig. 3c; P54.0|1026, ANOVA). However, the nature of those changes was strongly influenced by heat treatment (Fig. 3c). If we posit that differences in relative recovery between untreated and heat-treated samples reflect the presence and absence of predators, then the difference between these samples 614 represents the fitness against these predators. We can estimate fitness w as the ratio of a strain’s relative recovery in the presence or absence of predators (equation 3) (Fig. 3d), using the variability between the four replicates to calculate a variance (error bars represent 2 SD ). Here, LD906 has the highest fitness, meaning it had the greatest increase in relative abundance in the untreated sample relative to the heat-treated sample. Heat treatment does not release nutrients allowing growth While the decline in bacterial numbers over time in untreated samples is conventionally attributed to the action of predators (e.g. Fig. 1), one could propose that treatment with heat or anti-predator drugs releases nutrients that allow cell growth. If so, then one may observe differences in relative recovery because of differentially enhanced growth. While the strains LD906–910 are isogenic except for the region of the O-antigen-encoding rfb locus (Table 1), one may posit that those differences could be sufficient to utilize differentially any nutrients released by boiling. To test this hypothesis, we gathered four sets of soil samples (from two locations) and four sets of sediment samples (from two locations); samples were heat-treated as above, with an equal volume of water being added to soil samples before boiling. The environmental sample or water control (both buffered to pH 7) was inoculated to 107 c.f.u. ml21 with LD907. We monitored change in bacterial numbers in each sample or sterile water control in Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 Microbiology 162 Differential survival against predator communities 0.08 0.85 0.04 0.75 (e) 1.0 0.6 (L D 90 6) 1, 9, 12 (L D 91 0) 6, 7 (d) 1.4 Fitness, RU/RH 0.95 (L D 90 9) (L D 90 8) 3, 10 (L D 90 6) 1, 9, 12 (L D 91 0) 6, 7 (L D 90 9) 8, 20 (L D 90 7) 6, 8 (L D 90 8) 3, 10 (c) 0.12 1.2 0.8 0.4 0.2 0.65 (L D 90 6) 0 0 0.4 0.8 f40–U/f40–H 0.12 1, 9, 12 (L D 91 0) 6, 7 (L D 90 9) 8, 20 (L D 90 7) 6, 8 3, 10 (L D 90 8) 0 3, 10 6, 8 8, 20 6 1, ,7 9, 12 Fraction each serovar, f40 0.1 0.60 0 0.4 8, 20 0.66 0.06 1.0 (L D 90 7) 0.12 4.0 6, 8 0.72 Relative recovery, f40/f0 (b) 0.18 Fitness, RU/RH Fraction each serovar, f0 (a) Fig. 3. Differential fitness of Salmonella serovars (LD906–910) against heat-labile agents in natural environments. (a) Recovery of Salmonella serovars from sediments after 0 h incubation ( f0). White bars, heat-treated; grey bars, untreated. P50.18 (ANOVA). Values for LD906 are plotted using the right axis. (b) Recovery of Salmonella serovars after 40 h incubation ( f40). P54.0610–6. (c) Relative recovery was calculated as the ratio of fractions of cells recovered. P51.1610–9. (a–c) Grey bars, treated; white bars, untreated. (d) Fitness was calculated as the ratio of relative recovery for each serovar without (U) and with (H) heat treatment. Error bars represent 2 SD , which was calculated using the variances of each relative recovery coefficient. (e) Comparison of fitness and relative recovery. R 250.9972. eight replicate experiments. Two aliquots were removed from each replicate after either 0 or 48 h incubation at 25 uC; each aliquot was independently diluted and plated to estimate c.f.u. The mean change in cell numbers between 0 and 48 h was then computed for each of the eight replicates grown in either environmental extract or water. The significance of any difference between environmental sample and water was determined by a t-test. This was a conservative test in two ways. First, we were testing for any growth advantage for boiled environmental samples compared with water, which measured any nutrients released by boiling as well as any present in the sample prior to boiling. Second, we were measuring growth in an inoculum of 107 c.f.u. ml21, instead of the 108 c.f.u. ml21 added during effective competition experiments (e.g. Fig. 3); with 10-fold fewer cells being added, each cell has 10 times the potential nutrients available for growth. http://mic.microbiologyresearch.org As shown in Table 2, even under these conditions designed to maximize our ability to detect cell growth, bacterial cells did not show any growth difference between heat-treated environmental samples and water controls, neither of which showed any significant gain or loss of cell numbers during the course of the experiment. Therefore, we conclude that there was no significant release of nutrients due to heat treatment of either soil samples or sediment samples under our experimental conditions. Relative survival is affected by drugs targeting eukaryotic predators Heat treatment could affect numerous classes of predators, including bacteriophages, bacteria, unicellular eukaryotes such as amoebae, and multicellular eukaryotes such as nematodes. The effect of antiprotozoal and antihelminthic drugs in increasing overall survival (Fig. 1a) suggests that Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 615 A. Atzinger, K. Butela and J. G. Lawrence Table 2. Test of growth differences in buffered water or environmental samples Environmental sample* Environment Homewood Cemetery Phipps (Fern Room) Phipps (Desert Room) Panther Hollow Lake All locations Water control m s2 N m s2 N t df P 0.164 0.380 20.205 20.033 0.075 0.457 0.239 0.402 0.284 0.457 14 15 15 15 59 0.227 0.403 20.193 0.081 0.134 0.467 0.315 0.302 0.137 0.467 15 16 15 15 61 1.124 0.504 0.253 0.518 0.947 25.8 20.9 27.9 28.0 112.4 0.272 0.620 0.802 0.609 0.346 *Growth is reported as log2(N40/N0), where N0 and N40 are cell counts after 0 and 40 h incubation, respectively. Mean values are shown for replicate experiments for each environmental site paired with water controls using the same bacterial inoculum. eukaryotic predators are primary consumers of bacteria in this system. To examine the effects of antiprotozoal and antihelminthic drugs on relative survival of Salmonella, five fluorescently tagged, antigenically distinct but otherwise isogenic strains of Salmonella were competed within sediment samples collected from Panther Hollow Lake (Fig. 4); three replicate experiments were performed. Two treatments were used to attenuate predators: either the sample was heat-treated before equilibration to 25 uC, or the sample was exposed to the antiprotozoal drug metronidazole. Drugs used to attenuate eukaryotic predators do not affect growth or recovery of any genotype of Salmonella in a laboratory setting (Figs S2 and S3). The mixture of Salmonella was introduced and cells were recovered by antibody binding after either 0 or 30 h of incubation; cells were enumerated by flow cytometry. As expected, the fractions of each cell type were the same between untreated, heat-treated and drug-treated sediment at the start of the experiment, but were significantly different between untreated and heat-treated or drug-treated sediment after 30 h incubation. As expected from Fig. 3(c), relative recovery differed strongly between untreated and heat-treated samples (Fig. 4a; P59.3|1026, ANOVA). Notably, relative recovery also differed strongly between untreated and drug-treated samples (Fig. 4a; P51.2| 1025, ANOVA), indicating that the addition of metronidazole also allowed the survival of strains that were otherwise rapidly removed in the untreated sample. We can calculate fitness in the face of predation by comparing relative recovery in untreated versus heat-treated samples, or by comparing untreated versus drug-treated samples. As seen in Fig. 4b–d, the use of heat treatment or metronidazole treatment resulted in similar patterns of relative survival, indicating that both treatments attenuated predators in similar ways (Pearson R50.794). Metronidazole inhibits anaerobic amoebae, suggesting that amoebae affect relative survival of antigenically distinct Salmonella. This result was recapitulated in other soil and sediment samples (Fig. S4). In all cases, fitness in the face of predation was similar when using either heat or metronidazole to 616 incapacitate predators in either soil (Fig. S4a) or sediment samples (Fig. S4b). In addition, treatment with tetramisole and ivermectin also led to differential survival of antigenically distinct Salmonella (Fig. S4c); both ivermectin and tetramisole are broad-spectrum antihelminthic drugs, suggesting that nematodes both are major predators of bacteria in this environment and can mediate differential survival among antigenically distinct strains. Interestingly, the two classes of anti-eukaryote drugs led to similar fitness profiles (Fig. S4c), suggesting that amoebae and nematodes may share feeding preferences, or that these drugs both target a single class of predator. Nematodes discriminate between antigenically distinct strains of Salmonella To test the hypothesis that nematodes differentiate among prey, we grew a mixture of two antigenically distinct strains of Salmonella on solid media in the presence or absence of C. elegans. As expected, the fractions of each strain did not differ between plates bearing or lacking nematode predators at the start of the experiment (Pw0.05 at 0 h, Fig. 5). However, the ratios of bacterial strains differed after 4 days growth with or without C. elegans (Pv1023 at 96 h, Fig. 5). When competing WT strains SARB2 (O : 3,10) and SARB36 (O : 6,8), SARB2 increased abundance to a greater degree in the absence of predators, suggesting that C. elegans preferentially consumed that serovar. These results were validated in a replicate experiment (Fig. 5). To determine if C. elegans was discriminating based on the O-antigen, we competed near-isogenic strains HW31 (O : 1,4,12) and HW19 (O : 6,7), which differ only at their O-antigens and at the phs marker gene. In two replicate experiments, strain HW31 decreased in relative abundance in the absence of predators (Fig. 5), suggesting that C. elegans consumed it preferentially. Taken together, these data suggest that the action of antihelminthic drugs in conferring differential survival of Salmonella was consistent with nematodes differentially preying upon antigenically distinct strains of Salmonella. We discounted any role of nematode chemotaxis (Moens et al., 1999; Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 Microbiology 162 Differential survival against predator communities Relative recovery, f30/f0 (a) 1.6 1.2 0.8 0.4 0 3,10 (LD908) (b) 1.2 6,8 (LD907) 8,20 (LD909) (c) 6,7 (LD910) 1,9,12 (LD906) (d) 0.4 Fitness (drug) 0.8 Fitness, RU/RH Fitness, RU/RH 1.0 0.6 LD 9 L D 08 9 L D 07 9 L D 09 9 L D 10 90 6 0 LD 9 L D 08 9 L D 07 9 L D 09 9 L D 10 90 6 0.8 0.4 0.2 0.5 0.8 Fitness (heat) 1.1 Fig. 4. Fitness of antigenically distinct strains of S. enterica (LD906–910) in natural environments calculated using different treatments to attenuate predators. Fitness was calculated after 30 h incubation in sediment collected from Panther Hollow Lake. (a) Relative recovery was calculated as the ratio of fractions of cells recovered after 0 or 30 h; predators were attenuated by heat treatment (grey bars) or exposure to metronidazole (white bars). Black bars, untreated. (b) Fitness was calculated as the ratio of relative recovery for each serovar without and with heat treatment. Error bars represent 2 SD , which was calculated using the variances of each relative recovery coefficient. (c) Fitness was calculated as the ratio of relative recovery for each serovar without and with treatment with metronidazole. (d) Comparison of fitness values calculated using either heat or metronidazole to attenuate predators. R50.794. Ward, 1973) toward particular serovars because the two serovars were well mixed and abundant on the plates; strains HW31 and HW19 are also nearly isogenic, and would not differentially produce chemoattractants (Chaisson & Hallem, 2012; Prot, 1980) or cAMP (Ward, 1973). Fitness of Salmonella against predation varies between natural environments We have shown previously that amoebae isolated from different environments show different feeding preferences http://mic.microbiologyresearch.org (Wildschutte et al., 2004), though predators isolated from the same environment show similar feeding preferences (Wildschutte & Lawrence, 2007). Those results predict that Salmonella should show differential fitness when introduced into different natural environments, facing communities of predators with different sets of collective feeding preferences. To test this hypothesis, we repeated these competition experiments across soil and sediment collected from different locations, as well as the intestinal tract of a vertebrate host (Fig. 6). In each case, four replicates were performed for each experimental regime. Predators were attenuated using either heat treatment, exposure to the Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 617 P = 9.9×10–4 40 20 Time (h): P = 0.36 P= P = 0.25 P = 0.46 SARB2 or HW31 (%) 60 3.1×10–5 P = 2.5×10–3 P = 4.5×10–6 A. Atzinger, K. Butela and J. G. Lawrence 0 96 0 96 SARB2 vs SARB36 0 96 0 96 HW31 vs HW19 Fig. 5. Differential survival of Salmonella serovars against predation by C. elegans. Two strains of Salmonella were co-plated with (black bars) and without (white bars) C. elegans; the fraction of each strain was determined at 0 (start) and 96 h (stop). Ratios of SARB2 and SARB36 were assessed by plating on MacConkey– xylose medium. Ratios of HW31 and HW19 were assessed by plating on Kligler iron agar. P values are shown for ANOVAs testing differences between replicates with and without predators. antiprotozoal drug metronidazole, or exposure to a cocktail of metronidazole, ivermectin and tetramisole. To begin, we expected Salmonella competed within replicate samples collected from the same location to show similar fitness values (e.g. Fig. 3); this was upheld for competitions performed in numerous environmental samples, indicating that the method was robust. Because the nature of the fluorescent tag did not affect fitness (Fig. 2a), we expected a set of strains with the same O-antigens, but tagged with different fluorescent proteins, to show the same sets of fitnesses. This was tested using strains LD906–910 and strains LD911–916. Here, congruent fitness values (Pearson correlation R50.945) were obtained for sets of antigenically distinct, but otherwise isogenic Salmonella strains tagged with different sets of fluorescent proteins in experiment 3 Apr (Fig. 6). We also expected samples collected from the same location, but exposed to predators for different periods of time, to show congruent fitness patterns. This was observed in experiments 8 Aug, 2 Feb and 15 Mar (Fig. 6). As predicted by Fig. 4, we expected similar fitnesses for strains competed in experiments using one environmental sample wherein predators were attenuated using either heat or metronidazole; this is seen in experiments 3 Jul, 21 Sep, 2 Feb and 15 Mar (Fig. 6). These results bespeak the robustness of the data, as fitness values were strongly correlated, with a mean Pearson correlation of 0.748 in experiments at the same location using the same treatment for different times, or different treatments for the same time (Table 3). 618 However, fitness values for antigenically distinct, but otherwise isogenic, Salmonella strains were very different when they were competed in samples drawn from different locations (mean Pearson correlation 0.030; Fig. 6, Table 3), or from the same location at different times (mean Pearson correlation 20.091; Fig. 6, Table 3). The samples collected from the same pond location, but at different times, showed very different temperature profiles and different collections of leaf litter; therefore, it was not unexpected that predator communities would be quite different, leading to different relative fitness of Salmonella strains facing those predators. Taken together, these data demonstrate that communities of predators show an aggregate feeding preference. Predator identities are unclear, even though they share feeding preferences Soil and sediment environments harbour many predators of bacteria, including amoebae, ciliates, bacteriophage, predatory bacteria and nematodes. Heat treatment, which would eliminate all of these predators, increased bacterial survival (Fig. 1a) and revealed differential survival of antigenically distinct strains in the face of the predator communities (Figs 3 and 6). Treatment with antiprotozoal as well as antihelminthic drugs also increased bacterial survival (Fig. 1a), and all methods of attenuating predators revealed congruent patterns of differential survival (Figs 4 and 6). One conclusion that can be drawn is that the action of heat treatment in increasing survival (Fig. 1a) did not operate by removing bacteriophages, as neither antiprotozoal nor antihelminthic drugs would affect bacteriophages. The similarity of effects between antiprotozoal and antihelminthic drugs suggests that either both protozoa and nematodes share feeding preferences, or that the drugs affect a class of predators susceptible to both classes of drug. Metronidazole targets anaerobic protozoa by interfering with ferredoxin-mediated H2 production (Edwards & Mathison, 1970; Quon et al., 1992). In contrast, antihelminthic drugs act on multicellular eukaryotes. Ivermectin binds to glutamate-gated chloride channels, increasing their permeability to chloride (Cully et al., 1994; Vassilatis et al., 1997); this leads to hyperpolarization. Tetramisole acts as an acetylcholine receptor agonist, disrupting acetylcholine receptors in muscles and neurons (Martin & Robertson, 2007). It is not clear what organism would respond to all compounds. While metronidazole could impact some bacterial taxa, neither tetramisole nor ivermectin is known to affect prokaryotes. Since these drugs have similar effects on attenuating predators (Fig. 5), we conclude that the antibacterial effects of metranidazole were not responsible for the fitness differences between Salmonella serovars in our experiments. We propose that the amoebae and nematode predators within specific environments share feeding preferences. The strains competed herein were isogenic save for the genes encoding the O-antigen biosynthetic proteins; therefore, we conclude that predators recognize prey by virtue of their O-antigen polysaccharides. Yet other structures in these environments, Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 Microbiology 162 Differential survival against predator communities 1.2 Relative fitness 1.0 0.8 0.6 0.4 0.2 0 Time (h) Treatment Date Location 1,9,12 3,10 6,8 8,20 6,7 24S1 24S2 H H 48 H 3 Apr 6 Jun 40 H 40 M 3 Jul 30 H 30 C 20 H 21 Sep 30 H 8 Aug Sediment A 30 H 24 Sep Sediment B 24 H 48 H 48 M 24 H 48 H 48 M 24 H 2 Feb 15 Mar 31 May Soil A Soil B Fish Fig. 6. Relative fitness of isogenic, antigenically distinct strains of S. enterica in natural environments. Fitness against predation was calculated by attenuating predator by either heat treatment (H), exposure to metronidazole (M), or exposure to a cocktail (C) of metronidazole, ivermectin and tetramisole. All strains were LD906–910 except for 3 Apr set S2, which used LD911–915. Sediments A and B were collected from Panther Hollow Lake and Homewood Cemetery Pond, respectively; soils A and B were collected from the Fruit and Desert Rooms, respectively, at the Phipps Conservatory. Error bars represent two SD s calculated via resampling. including intestinal mucins and decaying plant matter, present surface polysaccharides. These structures are not viable food sources for these predators, and attempts at their ingestion would be fruitless. If the predators recognized such structures as environmental attachment sites rather than as food items, then bacteria with surface antigens resembling those structures would be conferred high fitness. This model of molecular mimicry was discussed previously to explain the similarity in feeding preferences shared between amoebae of different phyla (Wildschutte & Lawrence, 2007). Does differential predation lead to diversifying selection and bacterial speciation? Different strains of a single bacterial species are often found in different environments. For example, different genotypes of E. coli are found associated with different species of Table 3. Correlation in relative fitness values for competition among Salmonella serovars Pearson correlation Data subset Mean Same experiment, same treatment, 0.748 different time point Same experiment, different 0.748 treatment, same time point Same location, same treatment, 20.091 different date Different experiment 0.030 http://mic.microbiologyresearch.org Deviation Count 0.175 16 0.186 12 0.574 7 0.543 944 mammalian hosts (Gordon & FitzGibbon, 1999; Gordon & Cowling, 2003) or even different age groups in the same host species (Gordon et al., 2005). In such cases, the differential distribution of genotypes may reflect physiological differences between strains that adapt them to each environment. Here we provide evidence that predator escape may also drive the differential distribution of natural isolates between environments. The relative contributions of predator escape and physiological adaptation in arbitrating the differential distribution of genotypes among environments is unclear. We suggest that antigenic variability may allow bacteria to persist in novel environments by assisting in predator escape. This would allow the accumulation of physiological adaptations to permit more effective competition. Thus antigenic variability may provide the keystone adaptations that allow bacterial invasion of novel environments and, ultimately, speciation. Here, diversifying selection would maintain high antigenic diversity at the O-antigen-encoding rfb operon, counterselecting recombination in the region; this has been observed in enteric bacteria, where sequence divergence is high in the region flanking the rfb operon (Butela & Lawrence, 2009; Milkman et al., 2003). Therefore, selection for antigenic diversity at the rfb operon would represent the first step in fragmented speciation (Retchless & Lawrence, 2007), where adaptation to different environments imparts genetic isolation to a locus within the bacterial chromosome. ACKNOWLEDGEMENTS We thank Jessica Madden, Bryan Goddard and Alexis Fitzgerald for technical assistance, and Lewis Jacobson both for supplying C. elegans Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 619 A. Atzinger, K. Butela and J. G. Lawrence and for assistance with their husbandry. This work was supported by National Institutes of Health grant GM077548 to J. G. L. Luna, G. M., Manini, E. & Danovaro, R. (2002). Large fraction of dead REFERENCES Martin, R. J. & Robertson, A. P. (2007). Mode of action of levamisole Baumann, P. (2005). Biology bacteriocyte-associated endosymbionts and pyrantel, anthelmintic resistance, E153 and Q57. Parasitology 134, 1093–1104. of plant sap-sucking insects. Annu Rev Microbiol 59, 155–189. Milkman, R., Jaeger, E. & McBride, R. D. (2003). Molecular evolution Boyd, E. F., Wang, F.-S., Beltran, P., Plock, S. A., Nelson, K. & Selander, R. K. (1993). Salmonella reference collection B (SARB): of the Escherichia coli chromosome. VI. Two regions of high effective recombination. Genetics 163, 475–483. strains of 37 serovars of subspecies I. J Gen Microbiol 139, 1125–1132. Butela, K. & Lawrence, J. G. (2009). Population Genetics of Salmonella: Selection for Antigenic Diversity. In Bacterial Population Genetics in Infectious Disease, pp. 287–319. Edited by A. Robinson, D. Falush and E. J. Feil. London: Wiley-Blackwell. Butela, K. & Lawrence, J. (2010). Population genetics of Salmonella: selection for antigenic diversity. In Bacterial Population Genetics in Infectious Disease, pp. 287–319. Edited by D. A. Robinson, D. Falush & E. J. Feil. Hoboken, NJ: Wiley-Blackwell. and inactive bacteria in coastal marine sediments: comparison of protocols for determination and ecological significance. Appl Environ Microbiol 68, 3509–3513. Moens, T., Verbeeck, L., de Maeyer, A., Swings, J. & Vincx, M. (1999). Selective attraction of marine bacterivorous nematodes to their bacterial food. Mar Ecol Prog Ser 176, 165–178. Nyström, T. (2002). Aging in bacteria. Curr Opin Microbiol 5, 596–601. Nyström, T. (2007). A bacterial kind of aging. PLoS Genet 3, e224. Peduzzi, P. & Schiemer, F. (2004). Bacteria and viruses in the water column of tropical freshwater reservoirs. Environ Microbiol 6, 707–715. Butela, K. & Lawrence, J. G. (2012). Genetic manipulation of Pernthaler, J. (2005). Predation on prokaryotes in the water column pathogenicity loci in non-Typhimurium Salmonella. J Microbiol Methods 91, 477–482. Prot, J.-C. (1980). Migration of plant-parasitic nematodes towards Chaisson, K. E. & Hallem, E. A. (2012). Chemosensory behaviors of plant roots. Rev Nématol 3, 305–318. parasites. Trends Parasitol 28, 427–436. Quon, D. V., d’Oliveira, C. E. & Johnson, P. J. (1992). Reduced Chibani-Chennoufi, S., Bruttin, A., Dillmann, M. L. & Brüssow, H. (2004). Phage-host interaction: an ecological perspective. J Bacteriol transcription of the ferredoxin gene in metronidazole-resistant Trichomonas vaginalis. Proc Natl Acad Sci U S A 89, 4402–4406. 186, 3677–3686. Reeves, P. (1995). Role of O-antigen variation in the immune Cully, D. F., Vassilatis, D. K., Liu, K. K., Paress, P. S., Van der Ploeg, L. H., Schaeffer, J. M. & Arena, J. P. (1994). Cloning of an avermectin- Retchless, A. C. & Lawrence, J. G. (2007). Temporal fragmentation of sensitive glutamate-gated chloride channel from Caenorhabditis elegans. Nature 371, 707–711. and its ecological implications. Nat Rev Microbiol 3, 537–546. response. Trends Microbiol 3, 381–386. speciation in bacteria. Science 317, 1093–1096. Rodrı́guez-Zaragoza, S. (1994). Ecology of free-living amoebae. Crit Dale, N. G. (1974). Bacteria in intertidal sediments: factors related to Rev Microbiol 20, 225–241. their distribution. Limnol Oceanogr 19, 509–518. Rønn, R., McCaig, A. E., Griffiths, B. S. & Prosser, J. I. (2002). Impact Edwards, D. I. & Mathison, G. E. (1970). The mode of action of of protozoan grazing on bacterial community structure in soil microcosms. Appl Environ Microbiol 68, 6094–6105. metronidazole against Trichomonas vaginalis. J Gen Microbiol 63, 297–302. Roszak, D. B. & Colwell, R. R. (1987). Survival strategies of bacteria in Gordon, D. M. & Cowling, A. (2003). The distribution and genetic the natural environment. Microbiol Rev 51, 365–379. structure of Escherichia coli in Australian vertebrates: host and geographic effects. Microbiology 149, 3575–3586. Sørensen, S. J., Schyberg, T. & Rønn, R. (1999). Predation by Gordon, D. M. & FitzGibbon, F. (1999). The distribution of enteric protozoa on Escherichia coli K12 in soil and transfer of resistance plasmid RP4 to indigenous bacteria in soil. Appl Soil Ecol 11, 79–90. bacteria from Australian mammals: host and geographical effects. Microbiology 145, 2663–2671. Trevors, J. T. (2010). One gram of soil: a microbial biochemical gene Gordon, D. M., Stern, S. E. & Collignon, P. J. (2005). Influence of the age and sex of human hosts on the distribution of Escherichia coli ECOR groups and virulence traits. Microbiology 151, 15–23. Grimont, P. A. D. & Weill, F.-X. (2007). Antigenic Formulae of the Salmonella Serovars, 9th edn. Paris: WHO Collaborating Centre for Reference and Research on Salmonella. Hahn, M. W. & Höfle, M. G. (2001). Grazing of protozoa and its effect library. Antonie van Leeuwenhoek 97, 99–106. Vassilatis, D. K., Arena, J. P., Plasterk, R. H., Wilkinson, H. A., Schaeffer, J. M., Cully, D. F. & Van der Ploeg, L. H. (1997). Genetic and biochemical evidence for a novel avermectin-sensitive chloride channel in Caenorhabditis elegans. Isolation and characterization. J Biol Chem 272, 33167–33174. Walk, S. T., Alm, E. W., Gordon, D. M., Ram, J. L., Toranzos, G. A., Tiedje, J. M. & Whittam, T. S. (2009). Cryptic lineages of the genus on populations of aquatic bacteria. FEMS Microbiol Ecol 35, 113–121. Escherichia. Appl Environ Microbiol 75, 6534–6544. Human Microbiome Consortium (2012). Structure, function and Ward, S. (1973). Chemotaxis by the nematode Caenorhabditis elegans: diversity of the healthy human microbiome. Nature 486, 207–214. identification of attractants and analysis of the response by use of mutants. Proc Natl Acad Sci U S A 70, 817–821. Jjemba, P. K. (2001). The interaction of protozoa with their potential prey bacteria in the rhizosphere. J Eukaryot Microbiol 48, 320–324. Lawrence, J. G., Butela, K. & Atzinger, A. (2013). A likelihood Watson, S. W., Novitsky, T. J., Quinby, H. L. & Valois, F. W. (1977). Determination of bacterial number and biomass in the marine environment. Appl Environ Microbiol 33, 940–946. approach to classifying fluorescent events collected by multicolor flow cytometry. J Microbiol Methods 94, 1–12. Whitman, W. B., Coleman, D. C. & Wiebe, W. J. (1998). Prokaryotes: Levin, B. R., Antonovics, J. & Sharma, H. (1988). Frequency- Wildschutte, H. & Lawrence, J. G. (2007). Differential Salmonella dependent selection in bacterial populations. Philos Trans R Soc Lond B Biol Sci 319, 459–472. survival against communities of intestinal amoebae. Microbiology 153, 1781–1789. 620 the unseen majority. Proc Natl Acad Sci U S A 95, 6578–6583. Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 Microbiology 162 Differential survival against predator communities Wildschutte, H., Wolfe, D. M., Tamewitz, A. & Lawrence, J. G. (2004). Protozoan predation, diversifying selection, and the evolution of antigenic diversity in Salmonella. Proc Natl Acad Sci U S A 101, 10644–10649. Wildschutte, H., Preheim, S. P., Hernandez, Y. & Polz, M. F. (2010). O-antigen diversity and lateral transfer of the wbe region among Vibrio splendidus isolates. Environ Microbiol 12, 2977–2987. http://mic.microbiologyresearch.org Wommack, K. E. & Colwell, R. R. (2000). Virioplankton: viruses in aquatic ecosystems. Microbiol Mol Biol Rev 64, 69–114. Yuste, E., Moya, A. & López-Galı́ndez, C. (2002). Frequency- dependent selection in human immunodeficiency virus type 1. J Gen Virol 83, 103–106. Edited by: R. Lan Downloaded from www.microbiologyresearch.org by IP: 88.99.165.207 On: Fri, 05 May 2017 18:26:55 621