INTRODUCTION

Top Abstract Introduction Materials and Methods References

Ralstonia solanacearum, a phytopathogenic bacterium that causes a lethal wilting disease of many plants, produces multiple virulence factors (21, 22, 43). Among these factors is an acidic high-molecular-mass extracellular polysaccharide (EPS1) (37), which is produced in copious amounts by R. solanacearum both in culture and in planta (2, 14, 32). EPS1 is important for both the rapid systemic colonization of tomato plants by the pathogen and the subsequent wilt symptoms (2, 14, 27, 42). The 18-kb eps operon, which contains genes likely involved in both biosynthesis and export of EPS1, has been cloned, and about the first half of the operon has been sequenced (14, 25). Expression of eps and several other virulence genes is controlled by a complex, environmentally responsive, regulatory network (24, 26, 43). Part of the network constitutes a novel autoinduction system that uses as the extracellular signal molecule 3-hydroxypalmitic acid methyl ester, which is partially responsible for the cell density- or idiophase-dependent expression of the eps operon observed in culture (10, 17, 18). Thus, eps transcription is low in cultures below 10⁷ cells/ml but increases 30- to 50-fold during the next four generations (9). The production of EPS1 in culture follows the same pattern (32).

The behavior of R. solanacearum in culture suggests that EPS1 production per cell should be low during the early stages of infection and colonization but then should greatly increase as the bacterial population increases later in pathogenesis. However, since the environment inside plants is unlike that in bacterial cultures, expression of virulence genes in R. solanacearum during pathogenesis might also be different. This type of differential expression occurs in the hrp (hypersensitive response and pathogenicity) gene cluster (29) in phytopathogenic bacteria, because most hrp genes are repressed in rich culture media but are induced in planta (29, 52). In almost all cases, there is no evidence that induction of hrp genes in planta is due to unique signals, since comparable gene induction can be achieved in vitro by adjusting the pH, osmolarity, and carbon and nitrogen sources of the medium (29, 52). However, Marenda et al. (31) recently reported that R. solanacearum is an exception, since maximum expression of its hrp genes is observed only during cocultivation of the pathogen with tomato or Arabidopsis thaliana cell suspensions. Although there was no evidence for a diffusible plant compound inducing hrp expression in R. solanacearum, plant-specific signal molecules do induce nodulation genes and modification of lipopolysaccharide in Rhizobium spp. (16, 48), virulence genes in Agrobacterium tumefaciens (45), and phytotoxin biosynthetic genes in Pseudomonas syringae pv. syringae (35).

Determining how eps and other virulence genes in R. solanacearum are regulated during infection, colonization, and symptom development might significantly affect how we view these important processes. Unfortunately, methods for quantifying pathogen gene expression in planta are crude. Most researchers inject artificially large numbers of the pathogen carrying a reporter gene fusion into leaf intercellular spaces or add them to plant cell cultures and soon after recover the bacteria for use in standard enzyme assays (1, 3, 44, 55, 56). This approach is unsuitable for studying R. solanacearum in planta, because we must examine gene expression in stems while the pathogen population increases during several days of pathogenesis. In addition, since the distribution of bacteria within stem tissue might be nonuniform, we need a method that can detect subpopulations of the pathogen that differentially express eps genes. Therefore, we felt it essential to quantify eps expression in single cells of R. solanacearum recovered from infected tomato plants.

Multiple methods have been developed recently to detect gene expression in single bacterial cells, but most are either insensitive, quite complex, or require uncommon equipment, and none appear to have been used to quantify gene expression. We report here our development of a quantitative immunofluorescence (QIF) method to measure gene expression in single bacterial cells. Although immunofluorescence (IF) has been used to study the subcellular localizations of proteins in gram-negative and gram-positive bacteria (20, 30), we found no reports of its use to quantify the amount of a target antigen in single cells. Application of QIF to R. solanacearum cells recovered from plants revealed that expression of the eps operon during pathogenesis was very similar to that in culture, suggesting that no special conditions or signals in planta regulate this process.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods References

Strains and growth conditions. R. solanacearum strains (Table 1) were grown routinely at 30°C in BG broth or agar (9) or on BGT agar, to which has been added 50 µg of 2,3,5-triphenyl tetrazolium chloride per ml. Escherichia coli strains were grown at 37°C in Luria-Bertani medium (34). To assess -galactosidase activity qualitatively, X-Gal (5-bromo-4-chloro-3-indolyl--D-galactopyranoside) was incorporated into BG agar at a final concentration of 48 µg/ml. The antibiotics used were kanamycin (50 µg/ml), nalidixic acid (20 µg/ml), spectinomycin (50 µg/ml), tetracycline (15 µg/ml), and trimethoprim (100 µg/ml).

Growth and inoculation of tomato plants. Tomato seeds (Lycopersicon esculentum Mill., cv. Marion) were germinated in vermiculite, and 10-day-old seedlings were transplanted into 4-in. plastic pots containing a mixture of composted pine bark and vermiculite (3:1) amended with lime and fertilizer. After 3 to 4 weeks in a greenhouse, the plants were transferred to a growth chamber (30°C day, 25°C night, 13-h photoperiod) and inoculated the next day. To assess the virulence of R. solanacearum AW1-19cis, a bacterial suspension was prepared and poured onto the soil as described by Saile et al. (42). To assess gene expression in planta, a single colony from a fresh culture on a BGT agar plate (with appropriate antibiotics) was transferred to 200 ml of BG broth containing antibiotics and the culture was shaken at 250 rpm. To prepare the inoculum, bacteria were harvested when eps expression was still very low (i.e., before the culture reached an optical density at 600 nm of 0.01, or <10⁷ cells/ml) by centrifugation at 8,800 × g for 15 min; the cell pellet was then suspended in sterile water and adjusted to approximately 10⁶ cells/ml by inspection of turbidity. Plants were inoculated by excising the first leaf above the cotyledon 0.5 to 1 cm from its base and applying a 10-µl droplet to the cut surface.

Recovery of bacteria from infected tomato plants. Plants were usually sampled on a daily basis for 5 days after inoculation, with nine plants from each treatment being selected at random for processing on each day. In the laboratory, the stub of each petiole was cut off flush with the stem and a 1.0-cm-long segment of each stem, centered on the base of the inoculated petiole, was excised and weighed. Each stem segment was cut into 8 to 10 thin slices rather than being pulverized to minimize the amount of plant debris present in the resulting bacterial suspensions. The slices from three stems were pooled, added to tubes containing 3 ml of ice-cold 10 mM potassium phosphate buffer (pH 7.0), and held on ice without agitation for 1 h to allow bacteria to be released from the tissue. The tubes were vortexed and the number of bacteria in the buffer was determined by plate counting. Recovery of bacteria into the buffer was estimated to be 30 to 50% of the total based on preliminary tests examining the number of bacteria remaining in the tissue slices. One milliliter of each sample was transferred to a microcentrifuge tube, washed one time with phosphate-buffered saline (PBS; pH 7.4), and suspended in PBS (10 to 500 µl) to give about 10⁸ cells/ml. The suspensions were either used immediately or, more often, frozen at 20°C until they were processed for IF.

IF procedures. Methods for fixation, permeabilization, and IF staining of E. coli (30) and Bacillus subtilis (20) cells were modified for use with R. solanacearum as follows. Cells from BG cultures were washed once with PBS and suspended in PBS, whereas suspensions of cells recovered from infected tomato plants were thawed on ice. Suspensions were vortexed to disrupt clumps of cells, and aliquots were mixed 1:1 with fixative (5% paraformaldehyde and 0.04% glutaraldehyde in PBS) at room temperature. After 30 min of fixation, the cells were washed three times with PBS by centrifugation and then suspended in 10 to 100 µl of GTE (50 mM glucose, 20 mM Tris-HCl [pH 7.5], 10 mM EDTA), containing 2 mg of lysozyme per ml. Ten-microliter samples were immediately distributed onto standard microscope slides, which had been treated previously with 0.1% (wt/vol) poly-L-lysine (Sigma), and allowed to air dry at room temperature for 30 min. Slides were immersed in 20°C methanol for 5 min and then in 20°C acetone for 30 s and allowed to dry. Slides were flooded with 1 ml of blocking solution (2% bovine serum albumin [BSA] in PBS [BSA-PBS]) and incubated for 30 min at room temperature, and the liquid was drained off. For IF labeling, cells were incubated with a 1:5,000 dilution in BSA-PBS of mouse monoclonal anti--galactosidase antibodies (Promega) overnight at 4°C and washed five times with PBST (PBS buffer containing 0.05% Tween 20 [J. T. Baker]). The cells were then incubated with a 10-µg/ml solution of secondary antibody in BSA-PBS (Oregon Green 514-conjugated goat anti-mouse immunoglobulin G; Molecular Probes) for 2 h at room temperature in the dark. The slides were then washed five times with PBST, and 10 µl of PBS or PBS containing 5 ng of propidium iodide (PI; Molecular Probes) per ml was used for mounting coverslips. Slides were kept in the dark for up to 4 h until microscopy was completed.

Microscopy and image processing. Slides were examined with a Nikon Eclipse TE300 microscope equipped with a 100-W mercury lamp for epifluorescent illumination. A Nikon 60× Plan Apo DIC objective lens (1.4 numerical aperture) was used for most observations and image acquisition. Cells labeled with Oregon Green 514-conjugated secondary antibodies were visualized with a Chroma FITC filter set (BP460-490 excitation filter, 510-nm dichroic filter, and BP512-550 emission filter; Chroma Technology Corp.). To minimize photobleaching during image acquisition, bacteria were subjected to blue light for less than 10 s total; the exposure time for image capture was always 0.25 s. Cells stained with PI were visualized by using a Chroma TRITC filter set (BP510-590 excitation filter, 580-nm dichroic filter, and LP 590 emission filter). Images were captured with a three-chip color, charge-coupled device camera (Optronics model CCD 72) mounted on the microscope and were digitized with a Matrox MVP-AT image capture board. Digital images were analyzed with the aid of IM-4000 software (version 3.46; Analytical Imaging Concepts). The data were filtered to remove fluorescent signals that were too small to be bacteria, the numbers of cells that exceeded each of seven predetermined signal intensity thresholds were counted, and finally the number of cells within each signal intensity range was calculated. About 500 cells per sample were analyzed to provide data for subsequent evaluation.

Other methods. Methods for DNA isolation, transformation of E. coli and R. solanacearum, and bacterial conjugation have been described previously (5, 7, 9, 10). Standard methods to quantify -galactosidase activity in whole populations of cells used either o-nitrophenyl -D-galactopyranoside or methylumbelliferyl -D-galactopyranoside as the substrates and are described in detail elsewhere (9, 34).

	RESULTS AND DISCUSSION

Construction of cis merodiploid strain AW1-19cis. To investigate eps gene expression in a fully virulent strain of R. solanacearum, it was necessary to introduce a properly regulated eps-lacZ reporter fusion without inactivating production of EPS1. Since none of our cloning vectors is stable in R. solanacearum in the absence of antibiotic selection, we constructed an eps/eps-lacZ cis merodiploid strain by integrating a cosmid containing the eps operon into the genome of a strain with an eps-lacZ fusion (see Materials and Methods). When screened on BG X-Gal plates, a large majority of the putative cis merodiploids were either nonmucoid or mucoid but LacZ and were discarded. Two mucoid LacZ⁺ colonies were selected and tested further in culture for wild-type growth rate, EPS1 production, high -galactosidase activity, and phenotypic stability with and without antibiotic selection. Only one of the strains met all of our criteria and was designated AW1-19cis. When maximally expressed, the eps-lacZ fusion in AW1-19cis codes for about 900 Miller units of -galactosidase activity. Interestingly, EPS1 and -galactosidase production in AW1-19cis were more stable in the absence of antibiotic selection. Virulence assays using a soil infestation procedure that requires natural infection of tomato plants via their roots showed that there was no significant difference in the time required for a wild-type strain and AW1-19cis to cause 50% wilt (results not shown). Culturing of bacteria recovered from infected plants consistently showed that only 1 to 3% of the population had resolved the cis merodiploid and become LacZ.

Development of a QIF assay. Rather than use -galactosidase as an enzymatic reporter system, it was used as a target protein for IF. We reasoned that the numbers of primary and secondary antibodies that bind to their targets should be reasonably consistent and so the fluorescence intensity of each cell should be directly proportional to the number of -galactosidase molecules encoded by the eps-lacZ reporter fusions. For our samples, the most important parameter in this process was the length of fixation prior to permeabilization and IF staining. We found that 30 to 60 min of fixation produced cells with maximal fluorescence after IF but that fixation for >3 h resulted in substantially less fluorescence. An appropriate fixation time should be determined empirically for each system, since it may vary with bacterial species, the culture conditions, the liquid the cells are suspended in, and the antigen stability.

Validation of the QIF assay. (i) Quantification is negligibly affected by photobleaching. Photobleaching of a fluorophore is a common problem and may potentially prevent accurate quantification of fluorescence intensity. Oregon Green 514 was selected as the fluorophore for QIF, since according to its manufacturer, it is more resistant to photobleaching than fluorescein. To test whether photobleaching would be a problem in our system, cells of AW1-19cis that were IF labeled with Oregon Green 514 were exposed to blue light for measured time intervals up to 10 min and the expression index after each exposure was calculated. The data showed that fluorescence intensity decreased less than 10% during the first 30 s (results not shown). Consequently, images were acquired with less than 10 s of total exposure to blue light (which included finding the field, focusing, and data capture), so photobleaching was considered to be negligible.

(ii) QIF can distinguish strains with low and high -galactosidase activities. To evaluate the sensitivity of the QIF assay and its ability to distinguish strains with different amounts of -galactosidase activity, we compared AW1-19cis to strain AW1-130, which has an eps-lacZ fusion that codes for about one-fifth as much -galactosidase activity as AW1-19cis has (9, 25). Bacteria were grown in BG broth to a density at which eps expression is maximal (optical density at 600 nm, 0.5; about 5 × 10⁸ cell/ml), and samples were removed for standard -galactosidase enzyme activity assays. Three samples were removed from each culture, each of which was used to prepare a separate slide with IF-stained cells, and five microscopic fields were analyzed on each slide. -Galactosidase enzyme activity in AW1-19cis was about 5-fold higher than in AW1-130, whereas the QIF expression index detected a 3.5-fold difference in the levels of activity of the two strains (Table 2), and the differences were statistically significant. The expression distribution (Table 2) revealed that cells were found in each of the three signal intensity ranges (low, medium, and high) for both strains (see also Fig. 3). For AW1-130, >90% of the cells were in the low range, as might be expected given its comparatively low -galactosidase activity. In contrast, although a majority of AW1-19cis cells were in the high range, large proportions of the population were in the medium and low ranges. We were surprised to find this degree of variation in expression within the population of AW1-19cis cells. However, heterogeneity of expression in bacterial populations that were expected to be synchronized has been observed in other systems (8, 12, 28, 50).

(iii) Fluorescence intensity is directly proportional to -galactosidase activity. For QIF to be accurate, fluorescence intensity and -galactosidase activity must be linearly related. To test this, we assayed pure cultures of AW1-19cis and a LacZ wild type, as well as cells mixed in three different ratios (9:1, 1:1, and 1:9). The -galactosidase activity in each mixture was determined by the standard assay, and QIF expression indices were determined with duplicate samples. The results (Fig. 1) showed that both -galactosidase activity and the expression index for each sample were very close to the predicted values (based on the results for the pure cultures) and that both showed a positive linear regression.

View larger version (25K): [in this window] [in a new window]   FIG. 1.   Linearity of QIF expression index. Pure cultures of AW1 (Lac) and AW1-19cis (Lac+) and three mixtures of the two strains (9:1, 1:1, 1:9) were assayed simultaneously for -galactosidase activity (expressed as nanomoles of 4-methylumbelliferone released per minute per 108 cells) and for their QIF expression indices. The means and standard deviations shown are from one of two comparable experiments. A strong linear relationship was observed between values for both -galactosidase activity (r2 = 0.986) and the expression index (r2 = 0.985), which agreed very well with the predicted values.

(iv) QIF can detect cell density-dependent eps expression in AW1-19cis in culture. The final step in validating our system was to check whether expression of the eps-lacZ fusion in AW1-19cis is cell density dependent and to ensure that QIF can detect this change. BG broth cultures were started with about 10⁵ cells/ml and monitored for both -galactosidase activity and the QIF expression index over time. Regardless of the assay method, eps expression was low until the culture reached about 10⁸ cells/ml but then increased dramatically in the next several generations (Fig. 2A). The expression distribution (Fig. 2B and 3) showed that at low cell density, when eps expression was low, >90% of the cells were in the low signal intensity range; the percentages of bacteria in the medium and high signal intensity ranges increased as expression increased at higher cell densities. Therefore, both the cis merodiploid strain and the QIF assay were deemed adequate to study regulated eps gene expression in planta.

View larger version (38K): [in this window] [in a new window]   FIG. 2.   Detection of cell density-dependent eps expression in AW1-19cis in culture. Samples were removed from a BG broth culture over time and simultaneously assayed for -galactosidase activity and by QIF. (A) Comparison of -galactosidase activities (expressed as nanomoles of 4-methylumbelliferone released per minute per 108 cells) and QIF expression indices. The means and standard deviations shown are from one of three comparable experiments. (B) Expression distribution shown as the percentages of cells in the low (L), medium (M), and high (H) signal intensity ranges for each sample point.

View larger version (18K): [in this window] [in a new window]   FIG. 3.   Digital images illustrating the result of threshold analysis. To illustrate how digital-image analysis allowed determination of the number of bacteria in each signal intensity range (low, medium, or high), two of the original image files (from samples at low and high cell densities) used for the results presented in Fig. 2 were converted to a tag image format. Thresholds similar to those used in the actual data analysis were set with Adobe Photoshop 4.01 (Adobe Systems, Inc.), each manipulated image was saved as a new file, and the figure was assembled.

Expression of eps in AW1-19cis in planta. To monitor eps expression during pathogenesis, inoculum was prepared from low-density cultures of AW1-19cis (<10⁷ cells/ml) so that eps expression was at a low, uninduced level (Fig. 2) and a small number of bacteria were applied to wounded petioles. The bacteria colonizing the nearby stem tissue were then recovered and analyzed by QIF. The results were similar in three separate experiments, so the data are shown for one experiment in which plants were inoculated with 4.7 × 10⁴ bacteria and sampled daily for 5 days. The number of R. solanacearum cells recovered each day increased rapidly, from 6 × 10² cells/g of stem tissue after 1 day to >10⁸ cells/g after 4 days, as the pathogen multiplied in planta. No other bacteria were detected in the infected stems. There were not enough cells for QIF on day 1, but starting with day 2 it is clear from the QIF expression index (Fig. 4A) that eps expression was low early in pathogenesis when the number of bacteria in the plant tissue was low but increased dramatically when the number of bacteria increased. The expression distribution of cells in planta (Fig. 4B) was also similar to that observed in culture (Fig. 2B).

View larger version (36K): [in this window] [in a new window]   FIG. 4.   QIF analysis of eps expression in AW1-19cis recovered from infected tomato stems. Bacteria were recovered from nine plants each day, pooled to give three samples, and analyzed by QIF. (A) Expression indices plotted versus the numbers of viable bacteria recovered per gram of stem tissue. The means and standard deviations shown are from one of three comparable experiments; the day each sample was taken is also shown. (B) Expression distributions for the same samples shown in panel A given as the percentages of cells in the low (L), medium (M), and high (H) signal intensity ranges.

Comparison of QIF to other methods. Interest in studying specific gene expression in individual bacteria has resulted in a variety of new methods. Most methods rely on reporter gene fusions that code for easily detectable enzymes (such as -galactosidase), light-emitting luciferases, or the green fluorescent protein from Aequorea victoria (53). In addition to the normal drawbacks of reporter fusions (e.g., altered expression of the target gene, imposition of a deleterious metabolic load, and requirement for exogenous compounds), there are additional problems when they are used to study single cells. For example, chromogenic substrates like those used for -galactosidase do not permit detection of low enzyme activity (23, 40). Fluorescent substrates are more sensitive (12, 47), but their propensity to penetrate into cells can vary with the strain and the growth medium (33, 38), or they must be artificially loaded (41), potentially resulting in substrate limitation that precludes accurate quantitation (41). Leakage of the released fluorophore from the bacteria can also be a problem (28, 33). Detection of single light-emitting bacteria requires a microscope equipped with a specialized low-light video camera (46, 49) that is unavailable at many institutions.