|Home | About | Journals | Submit | Contact Us | Français|
Vegetable seeds have the potential to disseminate and transmit foodborne bacterial pathogens. This study was undertaken to assess the abilities of selected Salmonella and enterohemorrhagic Escherichia coli (EHEC) strains to attach to fungicide-treated versus untreated, and intact versus mechanically damaged, seeds of alfalfa, fenugreek, lettuce, and tomato. Surface-sanitized seeds (2 g) were exposed to four individual strains of Salmonella or EHEC at 20°C for 5 h. Contaminated seeds were rinsed twice, each with 10 ml of sterilized water, before being soaked overnight in 5 ml of phosphate-buffered saline at 4°C. The seeds were then vortexed vigorously for 1 min, and pathogen populations in seed rinse water and soaking buffer were determined using a standard plate count assay. In general, the Salmonella cells had higher attachment ratios than the EHEC cells. Lettuce seeds by unit weight had the highest numbers of attached Salmonella or EHEC cells, followed by tomato, alfalfa, and fenugreek seeds. In contrast, individual fenugreek seeds had more attached pathogen cells, followed by lettuce, alfalfa, and tomato seeds. Significantly more Salmonella and EHEC cells attached to mechanically damaged seeds than to intact seeds (P < 0.05). Although, on average, significantly more Salmonella and EHEC cells were recovered from untreated than fungicide-treated seeds (P < 0.05), fungicide treatment did not significantly affect the attachment of individual bacterial strains to vegetable seeds (P > 0.05), with a few exceptions. This study fills gaps in the current body of literature and helps explain bacterial interactions with vegetable seeds with differing surface characteristics.
IMPORTANCE Vegetable seeds, specifically sprout seeds, have the potential to disseminate and transmit foodborne bacterial pathogens. This study investigated the interaction between two important bacterial pathogens, i.e., Salmonella and EHEC, and vegetable seeds with differing surface characteristics. This research helps understand whether seed surface structure, integrity, and fungicide treatment affect the interaction between bacterial cells and vegetable seeds.
Globally, fresh produce consumption has increased significantly in the last few decades because of the accrued health benefits (1). Since most fresh produce receives minimal processing and is often eaten raw, it can be a vehicle for transmitting foodborne pathogens (2). It is reported that in the United States, fresh produce was the most common vehicle for transmitting foodborne illness and 19% of the foodborne outbreaks and 24% of the illnesses that took place from 2004 to 2013 were associated with fresh produce (3). In general, enterohemorrhagic Escherichia coli (EHEC) and Salmonella enterica are the major bacterial causes of foodborne illnesses (4). Fresh produce that has been linked to outbreaks of Salmonella and EHEC infections includes lettuce and tomato, as well as alfalfa and fenugreek sprouts (5,–8).
Fresh produce contamination by pathogenic bacteria such as Salmonella and EHEC may occur at pre- or postharvest stages (9). Vegetable seeds can be a potential source and efficient vector of human and plant pathogens. Unsanitized vegetable seeds could lead to the contamination of fresh produce, especially sprouts (10). According to the U.S. Food and Drug Administration (11), most sprout outbreaks have been caused by seeds contaminated with bacterial pathogens before the sprouting process begins. Many pathogens can survive for months under the dry conditions used for seed storage. However, pathogen populations in the seeds are low and unevenly distributed, making them difficult to detect by routine seed testing. Furthermore, seed sanitation treatments have been shown to be ineffective in eliminating bacterial pathogens, especially those that are located in surface cavities (12) or internal vegetable seed tissues (13).
Several studies have investigated the behavior of Salmonella and EHEC on various vegetable tissues, such as stems and leaves (14, 15). However, only a few have assessed the attachment ability of bacterial pathogens on vegetable seeds (16,–18), particularly lettuce, tomato, and fenugreek seeds. Furthermore, most of the earlier studies concerning bacterial pathogens and vegetable seeds have focused on the efficacy of chemical treatments in reducing seedborne bacterial pathogen populations (19). Physical mechanisms of pathogen attachment to vegetable seeds have not been adequately addressed.
Numerous factors may affect bacterial interactions with seed surfaces. Seed coat characteristics have a significant impact on the levels of contamination by artificially inoculated bacterial pathogens (20). Barak et al. (14) observed fundamental differences between Salmonella and E. coli O157:H7 in the manner and degree of attachment to alfalfa sprouts. However, this differential attachment behavior has not been reported for alfalfa seeds or other types of vegetable or sprout seeds. This study was undertaken to investigate differences in the abilities of two human enteric pathogens (Salmonella enterica and EHEC) to attach to seeds of four different types of vegetables (alfalfa, fenugreek, tomato, and lettuce) and with different surface characteristics (fungicide-treated versus untreated and intact versus mechanically damaged).
Overall, the four Salmonella strains used in this study had similar abilities to attach to vegetable seeds (Table 1). However, among the EHEC strains, strain K4499 had a significantly higher (P < 0.05) attachment ratio (12.5%) (i.e., the number of attached cells relative to the number of inoculated cells) than the other three EHEC strains. The attachment ratios of H1730 (1.5%) and ATCC BAA-2326 (0.2%) were similar, but they were significantly lower than those of K4499 and F4546 (5.2%). The mean EHEC attachment ratio from the 2-log CFU/ml inoculation level (5.8%) was significantly higher (P < 0.05) than that from the 4-log CFU/ml inoculation level (3.9%), but Salmonella attachment ratios were similar from the two inoculation levels (Table 1). Lettuce seeds by unit weight (2 g) had the highest numbers of attached Salmonella cells (18.7%), followed by tomato (13.2%), alfalfa (11.3%), and fenugreek (6.0%) seeds (Table 1). A similar trend was observed with EHEC cells, except that there was no significant difference in the numbers of cells attached to alfalfa and fenugreek seeds. However, individual fenugreek seeds had the highest numbers of attached Salmonella and EHEC cells, followed by lettuce, alfalfa, and tomato seeds (Table 1). With regard to seed surface characteristics, significantly more Salmonella and EHEC cells attached to mechanically damaged seeds than to intact seeds (Table 1; P < 0.05). Salmonella attachment to untreated seeds was significantly higher (P < 0.05) than to thiram (dimethylcarbamothioylsulfanyl N,N-dimethylcarbamodithioate)-treated seeds. In contrast, the numbers of EHEC cells attached to thiram-treated and untreated seeds were not significantly different (Table 1).
In general, the Salmonella strains used in this study displayed greater attachment ratios than the EHEC strains (Fig. 1). However, the attachment ratio of K4499 was comparable to those of the four Salmonella strains. For most of the bacterial strains included in the study, the numbers of cells attached to vegetable seeds were higher than those that were recovered from seed rinsing waters, except for Salmonella enterica serovar Stanley and E. coli ATCC BAA-2326.
For EHEC at both inoculum levels and Salmonella at the 2-log CFU/ml inoculation level, bacterial populations recovered from each type of seed were similar to the original inoculum levels, except for tomato seeds (see Table S1 in the supplemental material). At the 4-log CFU/ml inoculation level, the number of inoculated Salmonella cells was also similar to those recovered from lettuce seeds, and the latter was not significantly different from the numbers of Salmonella cells recovered from alfalfa and fenugreek seeds (P > 0.05). For tomato seeds, the recovered pathogen levels were significantly lower than the original inoculum levels (Table S1). These data suggest that there was no significant bacterial growth during the 5-h attachment process at 20°C.
Although there were no significant differences in the mean attachment abilities of the four Salmonella strains used in this study, individual strains appeared to have a unique affinity to a specific type of vegetable seed (Table 2). Significantly more S. Baildon (14.4%) and S. Cubana (17.7%) cells attached to alfalfa seeds than cells of S. Montevideo (5.8%) and S. Stanley (7.4%). Additionally, significantly more S. Montevideo (17.0%) and S. Stanley (19.9%) cells attached to tomato seeds than cells of the other two Salmonella strains (9.6% for S. Baildon and 6.5% for S. Cubana). Salmonella Stanley displayed a relatively low attachment ability to fenugreek seeds (2.9%) in comparison to those of other Salmonella serotypes (from 6.2% to 7.9%). Among the EHEC strains, strain K4499 had the highest attachment ratio on every vegetable seed type tested, followed by strain F4546. The attachment ratios of H1730 and ATCC BAA-2326 were significantly lower than those of the other two EHEC strains (Table 3). More F4546 cells attached to lettuce (8.9%) and tomato (7.1%) seeds than alfalfa (3.1%) and fenugreek (1.9%) seeds, while more K4499 cells attached to lettuce seeds (28.7%), followed by tomato (11.0%) and alfalfa (6.0%) seeds. The numbers of K4499 cells recovered from alfalfa seeds were not significantly different from those that attached to fenugreek seeds (4.3%). Significantly more H1730 cells attached to lettuce seeds (3.7%) than to the other three vegetable seed types. Similar numbers of ATCC BAA-2326 cells attached to tomato (0.2%), fenugreek (0.3%), and lettuce (0.4%) seeds, but the number of cells that attached to lettuce seeds was significantly higher than that recovered from alfalfa seeds (0.1%) (Table 3).
The numbers of Salmonella cells recovered from damaged alfalfa and tomato seeds and S. Baildon and S. Cubana cells recovered from damaged fenugreek seeds were significantly higher (P < 0.05) than those from the corresponding seeds without mechanical damage (Table 4). Furthermore, the numbers of Salmonella cells recovered from damaged lettuce seeds and S. Montevideo and S. Stanley cells recovered from damaged fenugreek seeds were similar to those from their undamaged counterparts, except for lettuce seeds inoculated with S. Cubana (Table 4).
The numbers of EHEC cells recovered from damaged tomato seeds, of F4546, K4499, and H1730 cells recovered from damaged alfalfa seeds, of F4546, H1730, and ATCC BAA-2326 cells from damaged fenugreek seeds, and of ATCC BAA-2325 cells from damaged lettuce seeds were significantly higher (P < 0.05) than those from their corresponding intact seeds. For the rest of the samples, the numbers of EHEC cells on damaged and intact seeds were statistically similar, except for lettuce seeds that were inoculated with strain K4499 and plated on tryptic soy agar amended with 50 μg/ml nalidixic acid (NA-TSA) (Table 5). Damaged tomato seeds had the lowest number of attached bacterial pathogen cells, except for the samples inoculated with ATCC BAA-2326 (Tables 4 and and55).
Although, on average, significantly more Salmonella cells were recovered from untreated seeds than from thiram-treated seeds (Table 1), thiram treatment did not affect the attachment of individual bacterial strains to vegetable seeds, except for lettuce seeds inoculated with S. Cubana or S. Baildon (Table 6). Similarly, no significant difference in attachment ratio of EHEC cells was observed between thiram-treated seeds and untreated seeds, except for lettuce seeds inoculated with strain H1730 (Table 7).
Fenugreek seeds had a cuboid shape and were the largest of the four types of seeds used in this study. Lettuce seeds were long and thin and had granular structures on their surfaces (Fig. 2). Oval-shaped alfalfa seeds had relatively smooth surfaces, while round tomato seeds had a rough surface texture (Fig. 2). The scanning electron micrographs of the seed surfaces revealed exposed cavities on mechanically damaged fenugreek and alfalfa seeds and seed debris on mechanically damaged lettuce and tomato seeds (Fig. 2). More-detailed scanning electron micrographs of seed surface morphology revealed regular nodes and crevices on fenugreek seeds (Fig. 3A), crevices and irregular nodes on lettuce seeds (Fig. 3B), pubescent covering (fuzz) and crevices on tomato seeds (Fig. 3C), and slight cracks on alfalfa seeds (Fig. 3D).
The present study revealed that, on average, S. enterica had greater attachment ratios than EHEC. Similar observations were reported in several previous studies involving alfalfa and bean sprouts (14, 21), cantaloupe rind surface (22), and food contact surfaces (23). Difference in cell surface hydrophobicity between Salmonella and EHEC cells was believed to be one of the contributing factors for the observed phenomenon (22,–25). However, contradictory findings have also been reported. Takeuchi et al. (15) observed that more E. coli O157:H7 cells than S. Typhimurium cells attached to lettuce leaf surfaces. It is possible that cell surface hydrophobicity varies among bacterial species/serotypes as well as among individual strains within a bacterial species/serotype. Furthermore, intrinsic cell factors other than cell surface hydrophobicity may play important roles in the interaction between bacterial cells and contact surfaces (26).
Among the EHEC strains used in this study, E. coli O104:H4 ATCC BAA-2326 had the lowest attachment potential (Fig. 1). This bacterial strain was isolated from a fenugreek sprout-associated outbreak in Germany in 2011 which affected 3,842 people in a dozen countries (7). ATCC BAA-2326 evolved from an enteroaggregative E. coli (EAEC) strain that acquired the genes for Shiga toxin production (7). EAEC strains produce a large amount of extracellular polymeric substances (EPS) (27). While EPS enhance cell aggregation and biofilm formation, they usually impair initial bacterial attachment to contact surfaces (28, 29).
The numbers of the bacterial cells attached to vegetable seeds were generally higher than those that were recovered from seed rinsing waters, except for S. Stanley and E. coli ATCC BAA-2326 (Fig. 1). This suggests that more cells of these two bacteria were loosely associated with seed surfaces and more easily rinsed away. This observation was incongruent with a previous study by Barak et al. (14), who reported that more Salmonella and E. coli O157: H7 cells were recovered from sprout rinse water than from sprout samples themselves. This inconsistency may be due to specific bacterial strains and experimental conditions used in the two studies. Additionally, seeds and seed sprouts have different surface characteristics. Surfaces of seeds are rougher and have irregular shapes and larger surface areas, while sprouts are thinner, longer, and smoother and have smaller surface areas.
We observed that the attachment ratios of Salmonella cells ranged from 2.9% on fenugreek seeds contaminated with S. Stanley to 19.9% on lettuce seeds contaminated with S. Cubana and tomato seeds contaminated by S. Stanley. By comparison, EHEC attachment ratios ranged from 0.1% on alfalfa seeds contaminated with ATCC BAA-2326 to 28.7% on lettuce seeds contaminated with K4499 (Tables 2 and and3).3). Most of the previous studies reported bacterial pathogen levels on vegetable seeds in log CFU per gram, rather than in terms of attachment ratio. Fransisca et al. (18) exposed 300 g of alfalfa seeds to 300 ml of a 107 CFU/ml E. coli O157 culture for 2 min, followed by rinsing for 20 min with water, and 3.16 CFU/g of E. coli O157 cells were recovered from alfalfa seeds, which was equivalent to an attachment ratio of 0.01%. This value was lower than the attachment ratios observed for alfalfa seeds contaminated with EHEC in the present study (Table 3). Fransisca et al. (18) reported that each gram of seeds was exposed to 107 CFU of E. coli cells, while in our study each gram of vegetable seeds was exposed to 103 to 105 CFU of pathogen cells. In addition, different bacterial strains and attachment conditions were used in the two studies. Higher cell concentrations were also used as inocula in several other studies (30,–32), and as expected, the attachment ratios calculated from these studies were lower than what was observed in the present study.
In the current study, the inoculation level had no significant effect on Salmonella attachment ratios, indicating that the number of Salmonella cells attached to seed surfaces increased as inoculum concentration increased. However, EHEC attachment ratios from the 2-log CFU/ml inoculation level were significantly higher than those from the 4-log CFU/ml inoculation level. The precise reason for the observed phenomenon is not clear.
The surfaces of vegetable seeds are complex, and different types of seeds have different surface properties. As a result, they have different potentials to attract bacterial cells. Lettuce seeds by unit weight had the highest number of Salmonella and EHEC cells, followed by tomato, alfalfa, and fenugreek seeds (Table 1). According to scanning electron micrographs, lettuce seed surfaces have nodes and crevices, while tomato seeds are pubescent. These two types of seeds have a rougher surface than alfalfa seeds (Fig. 3), which explains, in part, why pathogen attachment ratios from alfalfa seeds were lower (Table 1). This observation is supported by previous studies that demonstrated that wrinkled or rough seeds were likely to harbor more bacteria and were more resistant to sanitizers than smooth seeds (17, 18). Additionally, in the current study, alfalfa seeds clumped together in an aqueous environment, which might have prevented bacterial cells from attaching to their surfaces. Among the four seed types used in the present study, fenugreek seeds by unit weight had the lowest Salmonella and EHEC attachment ratios. Fenugreek seeds are larger and heavier, and by unit weight they have smaller surface areas to interact with bacterial cells than other seeds. It is worth noting that the fact that the two types of sprout seeds used in this study did not have higher numbers of Salmonella and EHEC cells attached does not make them microbiologically safer. The increase in bacterial population during seed germination and sprouting can still pose significant threats to consumer health.
Salmonella and EHEC attachment ratios from intact tomato seeds were the lowest among all the seed types tested. In addition, the total numbers of bacterial pathogens recovered from tomato seeds were significantly lower than the numbers of inoculated bacterial pathogen cells. Commercial tomato seeds are processed through a natural fermentation process to eliminate chemical compounds that inhibit germination (32). Although tomato seeds are washed and dried after fermentation, residual fermentation compounds such as lycopene may be present in seed coat hairs, leading to an acidic seed surface (pH ca. 2.6; detailed data not shown). Furthermore, tomato seed fermentation mixtures may contain antimicrobial compounds that inhibit the attachment and viability of bacterial pathogens. Arkoun et al. (33) reported that lactic acid bacteria isolated from fermented tomato fruits produced a bacteriocin-like substance that had inhibited a variety of Gram-negative bacteria, including E. coli.
The numbers of bacterial cells that attached to mechanically damaged seeds were significantly higher than those that attached to intact seeds, with a few exceptions. The microtopography of tested seeds revealed that damaging seed coats resulted in cracks that can provide physical protection for bacterial cells. This may have made it more difficult for loosely attached bacterial cells to be rinsed away. This observation is in agreement with previous studies that showed that cells of L. monocytogenes and E. coli O157:H7 were more likely to attach to cut edges of cabbage and lettuce (15, 26, 34). One explanation for the recovery of higher numbers of bacterial cells from damaged seeds than from intact seeds is that mechanical damage may result in increased uptake of water and increased leakage of solutes from seeds (35), which may promote attachment and growth of bacterial pathogens. However, according to the results shown in Table S1, significant bacterial multiplication did not occur during the attachment experiment.
Although significantly more Salmonella and EHEC cells were recovered from untreated seeds than from treated seeds, treatment of vegetable seeds with thiram did not affect the attachment of individual bacterial strains to vegetable seeds. The only exception was lettuce seeds. According to the U.S. Environmental Protection Agency, thiram is a nonsystemic fungicide, seed protectant, and animal repellent. Thiram is applied to seeds prior to planting as a dust, wettable powder, or liquid. The chemical has antibacterial (36) and antifungal (37) activities; however, no significant difference in the levels of bacterial attachment was found between thiram-treated and untreated seeds in the present study. This could be due to the fact that most of the thiram might have been rinsed off during seed sanitization and subsequent rinsing.
In summary, we observed the interactions between two important bacterial pathogens, i.e., Salmonella and EHEC, and vegetable seeds with different surface characteristics. The study provides a better understanding of whether seed surface structure, integrity, and fungicide treatment affect bacterial interactions with vegetable seed surfaces. Bacterial pathogen attachment to vegetable seeds is the first step in colonization. The fate of pathogen cells after the initial attachment step also has a significant impact on fresh produce safety. A study is under way in our laboratory to assess whether Salmonella and EHEC cells could migrate from contaminated vegetable seeds to different tissues of seed sprouts and vegetable seedlings.
Four S. enterica strains, three E. coli O157:H7 strains, and one E. coli O104:H4 strain were used in this study (Table 8). The bacterial strains were stored at −70°C and recovered on tryptic soy agar (TSA) at 37°C for 16 h. The resulting cultures were purified on bismuth sulfite agar (BSA; Becton, Dickinson, Sparks, MD), sorbitol MacConkey (SMAC; Becton Dickinson, Sparks, MD) agar, and MacConkey (MAC; Becton, Dickinson, Sparks, MD) agar, respectively. Spontaneous mutant cells resistant to 50 μg/ml nalidixic acid (NA; MP Biomedicals, Santa Ana, CA) were selected and used in the experiments.
Thiram (dimethylcarbamothioylsulfanyl N,N-dimethylcarbamodithioate)-treated and untreated fenugreek (Trigonella foenum-graecum, cultivar unidentified), lettuce (Lactuca sativa cv. Iceberg), and tomato (Solanum lycopersicum cv. Roma) seeds, as well as untreated alfalfa (Medicago sativa, cultivar unidentified) seeds, were obtained from a commercial source (Otis S. Twilley Seed Co. Inc., Hodges, SC) and stored at 10°C within a month before use. Commercial alfalfa seeds were treated with thiram 75 WP wettable powder fungicide (Chemtura, Pekin, IL) in-house at a rate of 1.8 g/500 g of seeds in accordance with instructions provided by Norac Concepts, Inc. (38). In order to mechanically damage seeds, 50 g of each seed type was blended in a 14-speed blender (Oster, Milwaukee, WI) for 30 s, and seed debris was removed using a sterilized, fine-mesh sieve (Walmart, Bentonville, AR; hole size, 0.05 mm). Thiram-treated intact, untreated intact, thiram-treated mechanically damaged, and untreated mechanically damaged seeds of alfalfa, fenugreek, lettuce, and tomato were used in the study.
The experiment involving bacterial attachment to seed surfaces was performed based on the method described by Barak et al. (14) and Darsonval et al. (39) with modifications. Two grams of each type of seed (described above) was placed in 50-ml centrifuge tubes (Fisher Scientific, Asheville, NC) and sanitized with 10 ml of a 20,000-ppm sodium hypochlorite solution (pH 6.8; Becton Dickinson, Sparks, MD) at room temperature for 15 min with gentle mixing. The seeds were then neutralized with 10 ml of Dey-Engley neutralizing broth (Becton Dickinson) for 10 min with gentle mixing and rinsed twice, each with 10 ml of sterilized deionized water for 1 min. An overnight culture of each Salmonella and EHEC strain grown in Luria-Bertani no-salt broth supplemented with NA (50 μg/ml) was diluted in sterilized water, and 20 ml of two concentrations (102 and 104 CFU/ml) of each inoculum was added to the centrifuge tubes with sanitized seeds. The precise inoculation levels were determined by plating 0.1 ml of appropriately diluted cell suspensions on TSA amended with NA. Vegetable seeds in the centrifuge tubes were agitated horizontally at 100 rpm in an orbital platform shaker (model 3520; Lab-Line, IL, USA) at 20°C for 5 h. The inocula were then decanted to sterilized test tubes, and seeds were rinsed twice, each with 10 ml of sterilized water for 30 s with gentle mixing. The rinse water from each seed type was collected into a sterilized test tube. Seeds were then soaked overnight at 4°C in 5 ml of phosphate-buffered saline (pH 7.4) to release attached bacterial cells. Each sample in the experiment was duplicated, and all experiments were conducted twice.
After being soaked at 4°C overnight, seed samples were vortexed at 3,200 rpm (Fisher Scientific, Asheville, NC) for 1 min. The resulting samples were 10-fold serially diluted, and appropriate dilutions of samples inoculated with Salmonella were plated on BSA. Those that were inoculated with E. coli O157 or O104 were plated on SMAC or MAC agar amended with NA, respectively. Additionally, all samples were plated on TSA amended with NA (NA-TSA). Bacterial populations in seed rinse water were also determined, as described previously. The ratio of the number of attached cells to the number of inoculated cells (attachment ratio) and the ratio of the number of the cells recovered from seed rinse water to the number of inoculated cells were both reported. The total populations of unattached cells in spent inoculum suspension, attached cells recovered from vegetable seeds, and loosely attached cells in seed rinse water were compared with the bacterial counts in the original inocula. A significant difference between the sum population and the cell count in the inocula indicates bacterial growth during each experiment.
To observe the surface morphology of dry vegetable seeds used in the study, scanning electron microscopy was performed according to the approach outlined in the user's manual. Each type of dry seed was mounted directly on stubs using double-sided adhesive tape and sputter-coated with gold using an SPI module sputter coater (model 11428-AB; Structure Probe, Inc., West Chester, PA). The surface morphologies of seeds were examined using a Zeiss 1450EP scanning electron microscope (Carl Zeiss, Inc., Thornwood, NY). Digital images (65× [Fig. 2] and 2,000× [Fig. 3]) were captured using SmartSEM (Carl Zeiss, Inc., Thornwood, NY).
To estimate differences among attachment ratios of each tested bacterial strain on vegetable seeds, Fisher's least significant difference test in a general linear model was used for separation of means based on a 95% confidence level using SAS (version 9.4; SAS Institute Inc., Cary, NC). The same statistical test was used to analyze the differences in pathogen attachment to seeds with differing integrities and fungicide treatments.
We thank John Shields for providing technical assistance with scanning electron microscopy, Norac Concepts, Inc., for donating the fungicide thiram, and Larry Beuchat and Mark Harrison for providing some of the bacterial strains.
This material is based upon work that is supported by the National Institute of Food and Agriculture, U.S. Department of Agriculture, under award number 2014-67017-21705.
Supplemental material for this article may be found at https://doi.org/10.1128/AEM.03170-16.