Kinetics and Strain Specificity of Rhizosphere and Endophytic Colonization by Enteric Bacteria on Seedlings of Medicago sativa and Medicago truncatula

Abstract

The presence of human-pathogenic, enteric bacteria on the surface and in the interior of raw produce is a significant health concern. Several aspects of the biology of the interaction between these bacteria and alfalfa (Medicago sativa) seedlings are addressed here. A collection of enteric bacteria associated with alfalfa sprout contaminations, along with Escherichia coli K-12, Salmonella enterica serotype Typhimurium strain ATCC 14028, and an endophyte of maize, Klebsiella pneumoniae 342, were labeled with green fluorescent protein, and their abilities to colonize the rhizosphere and the interior of the plant were compared. These strains differed widely in their endophytic colonization abilities, with K. pneumoniae 342 and E. coli K-12 being the best and worst colonizers, respectively. The abilities of the pathogens were between those of K. pneumoniae 342 and E. coli K-12. All Salmonella bacteria colonized the interiors of the seedlings in high numbers with an inoculum of 10² CFU, although infection characteristics were different for each strain. For most strains, a strong correlation between endophytic colonization and rhizosphere colonization was observed. These results show significant strain specificity for plant entry by these strains. Significant colonization of lateral root cracks was observed, suggesting that this may be the site of entry into the plant for these bacteria. At low inoculum levels, a symbiosis mutant of Medicago truncatula, dmi1, was colonized in higher numbers on the rhizosphere and in the interior by a Salmonella endophyte than was the wild-type host. Endophytic entry of M. truncatula appears to occur by a mechanism independent of the symbiotic infections by Sinorhizobium meliloti or mycorrhizal fungi.

Since 1995, there have been several reports of pathogenic Salmonella contaminating raw produce (1, 23). Salmonella outbreaks attributed to alfalfa sprout contamination have been reported in Finland and in several states, including California, Kansas, Missouri, Nevada, Oregon, and Washington (18, 23, 24). In each of these cases, the seeds were contaminated prior to germination. Various approaches to surface sterilize vegetables do not eliminate the enteric pathogens (2, 24, 25, 26). This result led to the discovery that these enteric pathogens reside within the interiors of apples, lettuce, and alfalfa sprouts (4, 11, 16, 26). Gandhi et al. (11) showed that a Salmonella Stanley strain could reside within the interiors of alfalfa sprouts to a depth of 18 μm from the surface. No cells were detected on the surface.

Nothing is known about the genetics of the invasion of a plant's intercellular spaces by bacteria. It is not known whether this process is regulated at the gene level by either the plant host or the bacterium. That is, do genetic determinants in the host and/or the bacterium influence endophytic colonization? If strains of Salmonella differ in their abilities to colonize the plant host, it is reasonable to assume that genes within the bacteria can affect endophytic colonization. Similarly, if host genotypes of a given species differ in their abilities to be infected by a specific strain of Salmonella, it can be reasonably assumed that there are genetic determinants in the hosts that regulate infection by endophytes.

Similarly, nothing is known about the kinetics of endophytic invasion by bacteria. For example, how many bacterial cells must be present in the inoculum for plant infection? Does the extent of endophytic colonization by the bacterium increase with increasing inoculum size? Is the ability to colonize the rhizosphere correlated with the invasion of the plant interior? How many salmonellae are found in the interiors of alfalfa seedlings, and how is this related to inoculum size? In this work, the kinetics of endophytic colonization by several Salmonella strains was studied. Most of the Salmonella strains used here were derived from salmonellosis outbreaks associated with alfalfa sprouts. Rhizosphere colonization by these bacteria was also assessed in this work.

Another important unanswered question is whether the process of infection by endophytic bacteria is related to the infection of alfalfa roots by nitrogen-fixing root nodule bacteria or mycorrhizal fungi. This question encouraged us to examine the endophytic colonization of a mutant of Medicago truncatula that is not infected by Sinorhizobium meliloti or Glomus sp. (5). This mutant is called dmi1. This experiment also addresses the question of whether host genotype matters in endophytic colonization. Examination of the colonization patterns within such host mutants may lead to effective strategies to prevent the entry of bacteria into many foods. An M. truncatula mutant was chosen for this work since this plant is a model legume that is closely related to alfalfa. In addition to its role as a model plant, M. truncatula is used in agriculture as a forage legume called annual medic. In addition, there are several mutants of M. truncatula available that are defective in plant-microbe interactions, particularly symbiotic associations.

The mode of entry of these human pathogens into plants is unknown. The nitrogen-fixing endophytes of sugarcane and other plants, Gluconacetobacter diazotrophicus and Herbaspirillum seropedicae, colonize lateral root junctions in high numbers, making these junctions a likely site of plant entry (13, 14, 20). These bacteria also enter the xylem and may infect the stem cortex through the xylem (13, 14). Here the ability of enteric bacteria to colonize lateral root cracks and the plant's vascular tissue was examined.

MATERIALS AND METHODS

Bacterial strains and inoculum preparation.

The bacterial strains used in this work are listed in Table 1. The organisms were cultured on solid Luria-Bertani (LB) medium (pH 7.3; Difco Laboratories, Detroit, Mich.) at 37°C for 24 h. Tetracycline (10 μg/ml) and ampicillin (50 μg/ml) were added for plasmid maintenance following transformation with a stably maintained plasmid that expresses green fluorescent protein (GFP) in culture and in planta (see below). Bacteria were suspended in phosphate-buffered saline (PBS; pH 7.2) to make concentrations of 10⁸ CFU/ml for inoculation. The bacterial suspensions were then diluted to provide inoculum levels of 10⁷, 10⁶, 10⁵, 10⁴, 10³, 10², 10¹, and 10⁰ CFU/seedling.

TABLE 1.

Bacterial strains used in this study

Strain	Comment	Reference or source
K. pneumoniae 342	Endophyte isolated from maize	7
S. enterica serovar Cubana strain H7976	Isolated from patient infected with S. enterica during alfalfa sprout-associated outbreak	Larry Beuchat, Center for Food Safety, University of Georgia
S. enterica serovar Infantis strain H3517	Isolated from patient infected with S. enterica during alfalfa sprout-associated outbreak	Larry Beuchat, Center for Food Safety, University of Georgia
S. enterica strain 8137	Isolated from patient with salmonellosis from alfalfa sprout contamination in Wisconsin	Wisconsin State Hygiene Laboratory
S. enterica serovar Typhimurium strain ATCC 14028	Clinical isolate used in many labs to examine Salmonella pathogenesis	American Type Culture Collection
E. coli K-12 strain MG1665	E. coli type strain	3
E. coli O157:H7 strain F4546	Isolated from patient infected with E. coli O157:H7 during outbreak associated with alfalfa sprouts	Robert Mandrell, Western Regional Research Center, USDA Agricultural Research Service, Albany, Calif.

GFP expression and plasmid construction and stability.

The plasmid gfpmut2 (9) contains a mutated GFP (GFP-S65T, V68L, S72A) which under the control of the promoter lac expresses a highly fluorescent GFP in Escherichia coli K-12 MG1665. Since many of the Salmonella strains used in this work are ampicillin and kanamycin resistant but tetracycline sensitive, a tetracycline resistance cassette was inserted into gfpmut2 as follows. A 2.0-kb HindIII fragment containing a tetracycline cassette from pKRP12 (19) was inserted into the HindIII site of gfpmut2, yielding pGFPTc2. Competent Klebsiella, Salmonella, and E. coli cells were transformed with pGFPTc2 by the calcium chloride method described by Sambrook et al. (22), which was slightly modified to use 50 mM rather than 100 mM CaCl₂. Transformed clones containing pGFPTc2 were directly selected as yellow-green colonies on solid LB medium containing ampicillin (50 μg/ml) and tetracycline (10 μg/ml). Only the GFP-labeled strains were used in this study. Thus, the GFP-tagged strains are referred to by their original names.

Plasmid stability in Salmonella enterica and E. coli strains was determined by growing those strains in LB liquid medium without selection pressure for over 24 generations (doubling times) and then plating a dilution series on both LB agar and LB agar with ampicillin and tetracycline. Four replicates were included for each treatment. The colonies on the plates were counted, and the ratio of colonies from plates with antibiotic selection pressure to those from plates without antibiotic selection pressure was determined.

Surface sterilization, germination, and inoculation of seeds. (i) Medicago sativa cv. CUS101.

Alfalfa seeds were obtained from J. W. Jung Seeds Co. (Randolph, Wis.). Alfalfa seeds (0.5 g per bacterial treatment) were surface sterilized by immersing the seeds in 70% ethanol for 5 min followed by three washes with sterile water. Seeds were then immersed in a 10% bleach solution for 20 min followed by three washes with sterile water. Surface-sterilized seeds were allowed to germinate on 0.5% agar LB plates overnight in the dark at room temperature. Seedlings with roots about 5 mm in length were transplanted into test tubes (18 by 150 mm) containing 10 ml of Jensen's N-free medium (15) with 0.45% agar. The next day, sprouts were inoculated with the appropriate inoculum level of the GFP-labeled strain of interest by dropping the cell suspensions directly to the seedling root area.

(ii) M. truncatula Gaertn. cv. A17 and dmi1.

Both wild-type M. truncatula and the dmi1 mutant, which is deficient in root nodulation by Sinorhizobium meliloti and/or mycorrhizal infection by Glomus, were kindly provided by D. Cook (University of California, Davis). Scarified seeds were surface sterilized by immersion in 70% ethanol for 2 min followed by three washes with sterile water, 5 min of immersion in 10% bleach, and another three washes with sterile water. Seed germination and bacterial inoculation were done as described above for alfalfa.

Plant culture conditions and harvesting.

All plants were placed in a growth chamber under a day and night cycle of 15 and 9 h, respectively, at 22°C. Alfalfa and M. truncatula plants were harvested 5 and 7 days after inoculation, respectively. Plants were removed from the growth medium, and the leaves were discarded. The fresh weights of the roots and hypocotyls were measured shortly after the removal of the leaves. There were four replicates for each treatment, with each replicate containing root and hypocotyl systems from four plants.

Determination of microbial populations in rhizosphere and inside surface-sterilized plant roots.

Plant tissues were suspended in a PBS solution containing 20% glycerol and subjected to a vortex at full speed for 1 min followed by sonication in a water bath for 1 min. The PBS solution was decanted, serially diluted in 10-fold increments, and cultured on solid LB medium with ampicillin (50 μg/ml) and tetracycline (10 μg/ml) to determine the microbial populations in the rhizosphere as described previously (21).

Roots and hypocotyls were gently dried, weighed, and immersed in 25 ml of a surface sterilization solution (1× PBS, 1% bleach, 0.1% sodium dodecyl sulfate, 0.2% Tween 20). Plant tissues were subjected to a vortex for 1 min followed by four washes with sterilized water. The samples were then crushed manually with a mortar and pestle for about 40 s. The homogenates were resuspended in 1 ml of PBS containing 20% glycerol. This suspension was serially diluted in 10-fold increments and cultured on LB plates supplemented with ampicillin (50 μg/ml) and tetracycline (10 μg/ml) in order to determine the microbial populations inside surfaced-sterilized roots and hypocotyls. To determine the efficacy of the surface sterilization, sterilized roots and hypocotyls were placed on LB plates containing ampicillin and tetracycline and incubated for 15 min prior to crushing. The roots were then removed, and the plate was further incubated at 37°C for 24 h as described by Gandhi et al. (11). In addition, the wash from the last root rinse was cultured to determine the efficiency of sterilization. In the very few cases in which the last wash yielded colonies with GFP, the number of those colonies was subtracted from the inside population, as described by Gyaneshwar et al. (12).

Slide preparation for microscopy and SCLM.

At least two seedlings from each of three independent inoculations were collected at 5, 7, and 9 days after inoculation. Samples of plant roots and hypocotyls were first surface sterilized and stored at 4°C for 2 h for the formation of the GFP chromophore (8). Root and hypocotyl pieces and hand-cut transverse sections of the hypocotyls and the maturation zones of the primary roots were mounted with an antifade kit (Prolong; Molecular Probes) and 0.25% agarose under a coverslip for microscopic observation. The coverslip was sealed with nail polish. Scanning confocal laser microscopy (SCLM) was done on a Bio-Rad MRC-1000 SCLM system attached to a Nikon Diaphot 200 inverted microscope. GFP-labeled cells were excited with the 488-nm laser line, and the autofluorescence from the plant tissues was excited with the 568-nm laser line. Images from both channels were collected sequentially in a z-series from 30 to 60 optical sections ranging from 0.5 to 2 μm in thickness. All microscopy experiments were repeated at least four times, and multiple samples were examined each time.

Data analysis.

Microsoft Excel 2000 was used to plot the endophytic and rhizosphere colonization data. Each plotted data point represents a mean and is surrounded by a representation of the standard error of the mean of the results from four replicates. Minitab 12 for Windows was used to perform Student's t tests to determine the significance of the differences between the means. Correlations between endophytic colonization and rhizosphere colonization were determined by regression analysis with Minitab 12. Where significant differences are described, the 95% confidence interval was used. Confocal photomicrographs were analyzed with the aid of Confocal Assistant 4.02 and Adobe Photoshop 5.0.

RESULTS

Assurance of endophytic colonization.

To ensure that the endophytic colonization numbers presented reflect only the number of cells within the interior of the plant tissues, a specific method of surface sterilization was developed. This sterilization method is sufficient to kill and/or wash away the surface bacteria while maintaining the survival of the interior bacteria. Three control procedures were performed to ensure that proper surface sterilization had been used in our experiments. First, root samples were examined by confocal microscopy for GFP-containing cells remaining on the plant surface after surface sterilization. No cells were ever observed by this approach. Second, no bacterial growth was observed on LB solid medium 2 days after the mounting of surface-sterilized roots. And third, bacterial growth was rarely observed when the last set of washes used to rinse the tissues was placed on LB medium. Colonies were observed in less than 1% of these washes. In those very few cases in which colonies were observed, the number of those colonies was subtracted from the final interior count of that particular sample. Based on the results of these three controls, the endophytic colonization numbers presented reflect only the number of cells within the interior of the tissues. Also, there were replicate inoculations for each treatment in each experiment. The standard error of the mean represented by each data point is presented in each figure.

Gandhi et al. (11) inserted the gfp gene into the chromosome of the Salmonella strain of interest to ensure the stability of gfp expression. Such stability is important to ensure accurate counting of cells on and within plant tissues. In this work, gfp was expressed on a plasmid that was stably maintained in the bacterial host for at least 24 generations. Plasmid-borne gfp was used in this work so that several strains could be labeled easily. As a result of this plasmid stability and the careful surface sterilization of plant tissues when needed, these results accurately reflect the numbers of enteric bacteria in the rhizospheres and the plant interiors.

Endophytic colonization of alfalfa by enteric bacteria.

At the four lowest inoculum levels, Klebsiella pneumoniae 342 colonized the interior of alfalfa in higher numbers than any other strain tested (Fig. 1A and Fig. 2A). At the four highest inoculum levels, only S. enterica serovar Cubana strain H7976 was able to match the level of endophytic colonization by K. pneumoniae 342 (Fig. 1A and Fig. 2A). Also, at the four highest inoculum levels, serovar Cubana strain H7976 colonized the interior of alfalfa in much higher numbers than all other strains except K. pneumoniae 342 (Fig. 1A and Fig. 2A). Among the other salmonellae, S. enterica serovar Typhimurium strain ATCC 14028, a clinical isolate used in many labs to examine Salmonella pathogenesis, colonized the interior in higher numbers than did S. enterica serovar Infantis strain H3517 and S. enterica strain 8137 over a range of inoculum levels from 10² to 10⁵ CFU per plant (Fig. 1A). At low inoculum levels, S. enterica strain 8137 was the poorest colonizer, but at high inoculum doses, this strain colonized the plant in numbers similar to those for all other Salmonella strains tested except serovar Cubana strain H7976 (Fig. 1A). At low inoculum levels, the levels of colonization by the two E. coli strains were between those of K. pneumoniae 342 and serovar Cubana strain H7976 (Fig. 2A). At the highest inoculum levels, the E. coli O157:H7 strain derived from an outbreak of diseases associated with alfalfa sprouts colonized the interior in higher numbers than did E. coli K-12 (Fig. 2A).

FIG. 1.

Numbers of GFP-labeled bacterial CFU recovered from interiors (A) and rhizospheres (B) of alfalfa seedlings 5 days after inoculation of 1.5-day-old seedlings with different inoculum levels. The data points represent the means and the bars represent the standard errors of the means of results from four replicate treatments. Four seedling samples (roots and hypocotyls) were taken from each replicate. SCH7976, S. enterica serovar Cubana strain H7976; Kp342, maize endophyte K. pneumoniae 342; SIH3517, S. enterica serovar Infantis strain H3517; S8137, S. enterica strain 8137; 14028, S. enterica serovar Typhimurium strain ATCC 14028; gfw, gram (fresh weight).

FIG. 2.

Numbers of bacterial CFU recovered from interiors (A) and rhizospheres (B) of alfalfa seedlings 5 days after inoculation of 1.5-day-old seedlings with different inoculum levels. The data points represent the means and the bars represent the standard errors of the means of results from four replicate treatments. Four seedling samples (roots and hypocotyls) were taken from each replicate. SCH7976, S. enterica serovar Cubana strain H7976; Kp342, maize endophyte K. pneumoniae 342; K12, E. coli K-12 strain MG1665; O157, E. coli O157:H7 strain F4546; gfw, gram (fresh weight).

Rhizosphere colonization by enteric bacteria.

K. pneumoniae 342 colonized the alfalfa rhizosphere at significantly higher levels than the other enteric bacteria tested at the four lowest levels of inoculum (Fig. 1B and Fig. 2B). At higher inoculum levels, K. pneumoniae 342 and serovar Cubana strain H7976 were very similar in their levels of colonization of the alfalfa rhizosphere (Fig. 1B and Fig. 2B). Over most inoculum levels, the Salmonella strains differed little in the levels of rhizosphere colonization (Fig. 1B). At levels of inoculum between 10² and 10⁵ CFU/plant, serovar Infantis strain H3517 was a relatively poor rhizosphere colonizer compared to most salmonellae tested (Fig. 1B). At the two lowest inoculum levels, E. coli K-12 was a better rhizosphere colonizer than E. coli O157:H7 (Fig. 2B). At the two highest inoculum levels, E. coli O157:H7 was a better rhizosphere colonizer than E. coli K-12 (Fig. 2B). No difference was observed between these two strains at the middle levels of inoculum (Fig. 2B). Across all inoculum levels, K. pneumoniae 342 was a better rhizosphere colonizer than either strain of E. coli (Fig. 2B).

Correlation between rhizosphere colonization and interior colonization of alfalfa.

There was a strong correlation between the abilities of all strains tested to colonize the rhizosphere and to colonize the interior of alfalfa. The correlation coefficients (r²) of alfalfa interior colonization and rhizosphere colonization by K. pneumoniae 342, serovar Cubana strain H7976, serovar Typhimurium strain ATCC 14028, serovar Infantis strain H3517, S. enterica strain 8137, E. coli O157:H7, and E. coli K-12 were 0.729, 0.891, 0.951, 0.949, 0.825, 0.803, and 0.017, respectively. Although this suggests a strong relationship between the two habitats, the strain most capable of entering the alfalfa plant, K. pneumoniae 342, has the second lowest correlation between rhizosphere colonization and endophytic colonization of the strains tested. In stark contrast to the other strains tested, E. coli K-12 showed no correlation between rhizosphere colonization and endophytic colonization. E. coli K-12 was a poor colonizer of the interior across all inoculum levels but a better-than-average colonizer of the rhizosphere at low inoculum doses.

Colonization of a symbiosis mutant of M. truncatula by serovar Cubana strain H7976.

Significantly higher numbers of serovar Cubana strain H7976 bacteria were able to colonize the interiors of the roots and hypocotyls of M. truncatula dmi1 than were able to colonize those of wild-type M. truncatula (Fig. 3J, K, and L; see also Fig. 5A). Furthermore, an even greater difference between the two host genotypes was found by examining rhizosphere colonization by serovar Cubana strain H7976. The numbers of serovar Cubana strain H7976 cells in the rhizospheres were approximately 100-fold higher on the roots of dmi1 than on those of the wild-type plant at all but the highest inoculum level (Fig. 4J, K, and L and Fig. 5B). The correlation between rhizosphere colonization and endophytic colonization was high for both genotypes, dmi1 (r² = 0.655) and the wild type (r² = 0.741). However, the correlation between these modes of colonization was much higher when serovar Cubana H7976 was inoculated on alfalfa (r² = 0.891).

FIG. 3.

Longitudinal section of alfalfa hypocotyls (A to H) and M. truncatula hypocotyls (J to L) and transverse section of alfalfa hypocotyls (I) showing colonization by the following GFP-labeled bacteria: E. coli K-12 (B), E. coli O157:H7 (C), K. pneumoniae 342 (D), S. enterica strain 8137 (E), S. enterica serovar Infantis strain H3517 (F), S. enterica serovar Typhimurium strain ATCC 14028 (G), and S. enterica serovar Cubana strain H7976 (H and I). (A) Uninoculated control. (J, K, and L) M. truncatula wild-type plant (J) and dmi1 mutant (K and L) inoculated with S. enterica serovar Cubana strain H7976. (B to L) Sections were visualized 9 days after inoculation. The inoculum level was 10⁴ CFU/plant. Arrows point to GFP-tagged bacterial cells. Bars, 50 μm.

FIG. 5.

Numbers of CFU recovered from interiors (A) and rhizospheres (B) of M. truncatula plant tissues 7 days after inoculation with different inoculum levels. SCH7976-WT, M. truncatula wild type inoculated with S. enterica serovar Cubana strain H7976; SCH7976-“Dmi1,” M. truncatula dmi1 mutant inoculated with S. enterica serovar Cubana strain H7976; gfw, gram (fresh weight).

FIG. 4.

Longitudinal section of alfalfa roots (A to I) and M. truncatula roots (J to L) showing colonization by the following GFP-labeled bacteria: E. coli K-12 (B), E. coli O157:H7 (C), K. pneumoniae 342 (D), S. enterica strain 8137 (E), S. enterica serovar Infantis strain H3517 (F), S. enterica serovar Typhimurium strain ATCC 14028 (G), and S. enterica serovar Cubana strain H7976 (H and I). (A) Uninoculated control. (J) M. truncatula without inoculation. (K and L) M. truncatula wild-type plant (K) and dmi1 mutant (L) inoculated with S. enterica serovar Cubana strain H7976. (B to L) Sections were visualized 9 days after inoculation. The inoculum level was 10⁴ CFU/plant. Arrows point to the lateral roots. Bars, 50 μm.

Lateral root crack colonization.

Cells of K. pneumoniae 342, S. enterica strain 8137, serovar Infantis strain H3517, serovar Typhimurium strain ATCC 14028, and serovar Cubana strain H7976 congregate at the lateral root emergence sites of alfalfa (Fig. 4). SCLM shows that these cells also colonize the interiors of alfalfa roots and hypocotyls. While S. enterica and E. coli strains showed the same patterns of colonization of lateral root emergence sites as K. pneumoniae 342, the numbers of S. enterica and E. coli cells at these sites appeared to be lower (Fig. 4). Serovar Cubana strain H7976 and K. pneumoniae 342 cells were observed in high numbers in hypocotyls by SCLM.

DISCUSSION

Several new aspects of the interaction between alfalfa and enteric bacteria are presented here. First, the number of cells that must be present in the inoculum for plant entry has been determined. With two of the four Salmonella strains used in this work, a single CFU in the inoculum was sufficient to cause invasion of the plant interior as determined 5 days after inoculation. This is a troubling result from a food protection perspective. This suggests that no contamination can be tolerated if the infection of alfalfa sprouts by certain salmonellae under the conditions used here is to be prevented. However, it is not known whether low contamination levels can cause significant invasion of the alfalfa apoplast under conditions in which sprouts are grown commercially. The presence of nonpathogenic bacteria on alfalfa seeds and in the nutrient solution and medium used to culture these plants commercially may exclude plant invasion by low numbers of salmonellae and/or may prevent the multiplication of these pathogens in planta.

Second, the rigorous surface sterilization procedure used here allowed the quantification of salmonellae within the plant. The numbers of salmonellae in planta increased with inoculum dose and reached levels as high as 4.5 log₁₀ per gram of tissue. These levels were nearly as high as those achieved by the Klebsiella endophyte. Thus, surface sterilization of tissue alone is not sufficient to eliminate the pathogen. The numbers of cells found in these plants are likely to be more than sufficient to cause disease if just one gram of tissue was available for human consumption, according to models by Latimer et al. (17). Gandhi et al. (11) and Charkowski et al. (6) determined the number of enteric pathogens on alfalfa seedlings. However, no surface sterilization was done in either study, so the numbers of surface and endophytic bacteria cannot be distinguished in these experiments. In this work, very thorough surface sterilization allows us to distinguish between rhizosphere colonization and endophytic colonization.

Third, four Salmonella strains were examined in this work while only one strain was studied by Gandhi et al. (11). The four strains differed greatly in their abilities to enter alfalfa roots and hypocotyls. Since strains can differ in their abilities to enter the same host genotypes cultured under the same environmental conditions, there must be genetic determinants in the bacteria than can affect endophytic colonization. These strain differences in endophytic colonization may occur through the presence of genes involved in this process or through a change in the expression of genes present in all strains. In either case, these data suggest that endophytic colonization by salmonellae is an active process and is not simply the passive diffusion of bacteria into the intercellular spaces of plants.

Fourth, the abilities of each of these strains to colonize the rhizosphere were assessed. This allows a comparison of rhizosphere colonization and interior colonization for each of the strains tested. A strong correlation was found between rhizosphere colonization and interior colonization for all strains except E. coli K-12, for which no correlation existed. The other six strains did exhibit a strong correlation between these two modes of colonization. However, it remains unclear whether the ability to colonize the rhizosphere enhances endophytic colonization. For example, the strain with the best ability to colonize the plant's interior, K. pneumoniae 342, showed the lowest correlation between rhizosphere colonization and endophytic colonization among the six strains with a strong positive correlation. In addition, the strain with the highest correlation, serovar Typhimurium strain ATCC 14028, was an average colonizer of both sites and is the only strain among the six strains that was not derived directly from plants or isolated from patients believed to have been infected through plant consumption. Nevertheless, it seems reasonable that a higher number of cells on the roots gives the strain a higher likelihood of multiple infection events, which would result in a higher interior colonization.

Fifth, a search was done with SCLM for bacterial colonization of lateral root cracks. Nitrogen-fixing endophytes colonize such cracks in high numbers (13, 14, 20). Micrographs presented here are representative of the extensive colonization of these regions on alfalfa roots by enteric bacteria. This finding suggests that these bacteria may enter the plant through these cracks. Alternatively, these root cracks may be a higher source of plant nutrients for the enteric bacteria than the rhizosphere. In addition, root hair deformation was not observed following inoculation with any of these bacteria. Thus, infection through root hairs, as in the infection of legume root hairs by rhizobia (10), is unlikely.

This work also expands on previous work in that a clear effect of the host genotype on endophytic colonization was observed. The dmi1 mutant of M. truncatula is defective in forming a symbiosis with either Sinorhizobium meliloti or mycorrhizae (5). This mutant is defective in a common and very early infection event in both symbioses. Prior to inoculation with serovar Cubana strain H7976, the expectation was that the dmi1 mutation would have no effect on endophytic colonization. A result that was thought to be less likely was that endophytic colonization would be impaired in the dmi1 mutant compared to that in the wild type. In contrast to both of these possibilities, both rhizosphere colonization and endophytic colonization were enhanced in the dmi1 mutant compared to those in the wild type, particularly at low inoculum doses. The effect on rhizosphere colonization was more pronounced. As the biochemical nature of the dmi1 mutation is unknown, the biological basis for these observations is not known. It appears that a blockage of infection by symbionts somehow reduces the barriers to endophytic infection at low inoculum doses. The higher level of endophytic colonization may be the effect of increased rhizosphere colonization. This alone may give more bacteria an opportunity to enter the plant root.

In summary, several important features of endophytic colonization by enteric bacteria have been characterized in this work. Very few cells must be present in the inoculum for colonization of the interior of the plant. Rhizosphere colonization and endophytic colonization are usually highly correlated. Strains of enteric bacteria can differ greatly from each other in their abilities to invade the plant interior and to colonize roots. As both bacterial and host genotypes were found to influence endophytic colonization, this infection is likely mediated by genetic determinants in both partners.