INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Many plant-associated bacteria have the ability to produce the plant growth regulator indole-3-acetic acid (IAA) (5, 9, 25, 33). IAA is involved in diseases caused by gall- and knot-forming bacterial species (33); however, its role in other bacteria remains undefined. It is unclear whether these bacteria produce IAA during colonization of plant surfaces and whether this metabolite is beneficial to the bacteria during their growth and survival in the phyllosphere. The production of IAA may enable bacteria to detoxify tryptophan analogues present on plant surfaces (15), to downregulate genes involved in plant defense responses (33), or to inhibit the development of the hypersensitive response by plants (26). We recently demonstrated that the ipdC gene, which encodes the indolepyruvate decarboxylase of Erwinia herbicola (Pantoea agglomerans) 299R and which is involved in the indolepyruvate pathway for IAA synthesis in this epiphytic strain (2), is osmoresponsive and plant inducible (3). We hypothesized that the secretion of IAA may modify the microhabitat of epiphytic bacteria by increasing nutrient leakage from plant cells; enhanced nutrient availability may better enable IAA-producing bacteria to colonize the phyllosphere and may contribute to their epiphytic fitness (1).

Few studies have attempted to determine the ecological significance of IAA production in pathogenic bacteria. Varvaro and Martella (31) showed that IAA-deficient mutants of Pseudomonas syringae pv. savastanoi, obtained by selection for resistance to -methyltryptophan, were reduced in their ability to colonize and survive on olive leaf surfaces. The survival of an -methyltryptophan-resistant IAA-deficient mutant of P. syringae pv. savastanoi in knots also was affected, its population declining more rapidly than that of the parental strain when inoculated alone into oleander leaf tissue (28). The importance of IAA production in bacterial colonization of bean leaves was also tested with the brown spot pathogen P. syringae pv. syringae and an IAA-deficient mutant derived by insertional mutagenesis (21). Although no difference in the survival of the parental and mutant strains on bean leaves was observed in the greenhouse, a small difference in their behavior was apparent in experiments conducted in a mist chamber (21). There have been no studies of the role of IAA production in plant-associated bacteria that do not cause disease.

IAA biosynthesis is not essential for bacterial growth and survival, since IAA-deficient mutants grow as well as their IAA-producing parental strain in vitro (2, 29). Large differences in the epiphytic behaviors of IAA-producing bacteria and isogenic IAA-deficient mutants consequently would not be expected. Even small contributions of IAA production to epiphytic fitness could account for the common presence of this phenotype in epiphytic bacteria (19). Measurements of changes in the ratio of two strains following coinoculation, a common approach in ecological studies, can allow the detection of even small differences in the competitive behaviors of two organisms. This approach can detect much smaller differences in behavior between closely related species than comparison of populations of these species when present singly in separate habitats (16). In this study, we tested the role of IAA in the epiphytic fitness of E. herbicola by comparing the relative changes in the population sizes of the parental and IAA-deficient mutant strains with time after their inoculation onto plants in both controlled and field environments.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Bacterial strains. The source and characteristics of E. herbicola 299R were described previously (1). E. herbicola 299XYLE is a derivative of strain 299R containing the xylE gene from Pseudomonas putida mt-2 (13) inserted into ipdC. The xylE gene encodes a 2,3-catechol dioxygenase catalyzing the conversion of catechol to 2-hydroxymuconic semialdehyde, a yellow compound (13). The xylE gene was inserted into ipdC by partial digestion of pMB2 with KpnI and ligation of xylE into the unique KpnI site of ipdC (2). The xylE derivative of pMB2 (pMB2A) was mobilized into strain 299R by triparental mating with helper plasmid pRK2013 (7). The ipdC gene of strain 299R was interrupted by homologous recombination of the xylE insertional derivative of ipdC into the chromosome to generate strain 299XYLE. In strain 299XYLE, xylE was fused to ipdC in the same transcriptional orientation as ipdC. Strain 299XYLE could therefore be distinguished from the parental strain by its ability to convert catechol to a yellow compound. Quantification of IAA and tryptophol (TOL) production in cultures of strains 299R and 299XYLE was performed by high-performance liquid chromatography as described previously (2).

Preparation of inoculum. The inoculum for IAA bioassays was prepared by growing strains to stationary phase in King's B medium at 27°C. The cells were then centrifuged, washed twice in potassium phosphate buffer (10 mM, pH 7.0) (KP buffer), resuspended at 10⁹ cells ml¹ in KP buffer, and used in bioassays. For the greenhouse and field studies, bacterial strains were cultured on Luria-Bertani agar supplemented with 100 µg of rifampin ml¹ (LBAR) for 24 h at 27°C. Bacterial cells were removed from the agar surface with a swab and suspended in KP buffer. The cell concentrations of the suspensions were determined turbidimetrically and adjusted by dilution in KP buffer. Appropriate volumes of the suspensions of the parental and mutant strains were combined to yield the following proportions of strain 299R to strain 299XYLE: 0:1, 1:0, 1:1, 1:10, and 10:1.

IAA root bioassay. The IAA root bioassay was performed as described by Loper and Schroth (20) with the following modifications. Radish seeds (Raphanus sativus L. cv. Comet) were washed by agitation in 100 ml of an aqueous solution containing 1% sodium hypochlorite and 100 µl of Tergitol for 10 min and then rinsed five times in 100 ml of sterile water. Ten seeds were placed at the top of each Seed-Pack Growth Pouch (Vaughan's Seed Company, Downers Grove, Ill.). The packs were wetted with 10 ml of an aqueous solution of an indole derivative or of a bacterial suspension in KP buffer (10⁹ cells ml¹); control treatments consisted of wetting the packs with water or KP buffer, respectively. Five replicate packs were prepared per treatment and incubated at room temperature in the dark for 4 days. Root length was recorded, 2-cm-long root tips (three from each pack) were crushed in 1.3 ml of KP buffer, and bacterial population size was estimated by plating on LBAR with a spiral diluter-plater (Spiral Systems, Inc., Cincinnati, Ohio).

Estimation of fitness in culture. Cells of both strain 299R and strain 299XYLE were cultured in minimal A medium (23) to stationary phase at 27°C, centrifuged, and resuspended in KP buffer. Appropriate volumes of the suspensions of the parental and mutant strains were combined to yield an equal number of cells of each strain, and a small aliquot of the mixed suspension was added to 10 ml of minimal A medium or minimal A medium containing 1.28% NaCl to yield a final total cell concentration of 5.5 × 10⁵ cells ml¹. Ten replicate cultures for each medium were prepared and incubated at 27°C on a rotary shaker. Culture aliquots were taken at mid-log phase and stationary phase of growth, and the cell concentration of each strain was estimated by dilution plating on LBAR. After 24 h of incubation, 10 µl of each culture was transferred to a new flask; the same procedure was repeated twice. The population size of each strain was estimated by a colony lift assay as described below.

Greenhouse study. Bacterial cells were applied to beans (Phaseolus vulgaris cv. Bush Blue Lake 274) at the first trifoliate leaf stage by immersing the plants (10 per pot) in a suspension of bacterial cells for 5 s. The cell concentration of the suspension was adjusted to 2 × 10³ cells ml¹ for each strain in single inoculations and in coinoculations at a ratio of 1:1. For coinoculations at a ratio of 1:10 or 10:1, the suspensions consisted of a mixture of 2 × 10³ and 2 × 10⁴ cells ml¹ of the appropriate strains. For each treatment, plants from five replicate pots were inoculated. The individual pots were covered with plastic bags to maintain a humid environment and were placed in a randomized complete block design on a bench in a greenhouse at 25°C. Two primary leaves were sampled randomly from each pot at regular time intervals during the 40 h following inoculation.

Field study. Field studies were conducted in a 5-year-old pear (Pyrus communis var. Bartlett) orchard located at the University of California Hopland Research and Extension Center in northern California. The trees in the plot were arranged in a randomized complete block design with four treatments and five replications. Each replicate treatment consisted of two adjacent trees in a row. One untreated guard tree separated each treatment area within a row, and two trees separated each treated experimental area from the adjacent trees in the orchard. The treatments consisted of an untreated control, strain 299R, and strain 299XYLE inoculated singly at 10⁶ cells ml¹ or the strains coinoculated in equal proportions (5 × 10⁵ cells ml¹ for each strain).

Estimation of bacterial population sizes. For the greenhouse study, individual bean leaves were placed in large tubes containing 20 ml of KP buffer, sonicated for 7 min, and vortexed vigorously. Appropriate dilutions of the suspension were plated on LBAR to estimate cell numbers. Each leaf was weighed to allow the bacterial population size to be normalized for the amount of leaf tissue.

Statistical methods. All statistical calculations were performed with SAS (version 6.04; SAS Institute Inc., Cary, N.C.).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Characterization of E. herbicola 299XYLE. Strain 299XYLE produced only 0.62 ± 0.08 (mean ± standard error) µg of IAA per ml when cultured for 48 h in minimal A medium supplemented with 0.02% tryptophan; TOL was undetectable. The IAA content of 299XYLE cultures was 14-fold lower than that of 299R cultures (data not shown). The xylE derivative of 299R therefore showed reduced TOL and IAA synthesis in vitro similar to that of other ipdC insertional mutants of 299R that were previously described, such as strain 299MX149 (2).

Effect of bacterial IAA production on roots. The root bioassay showed that IAA was the most potent inhibitor of growth among all indole compounds tested, reducing root elongation at concentrations of higher than 10⁷ M (Fig. 1). Indolepyruvate, indoleacetaldehyde, and tryptophan also were inhibitory, although only at concentrations 100-fold higher than those of IAA (Fig. 1). Strain 299R, which produces large quantities of IAA in culture, caused a 64% inhibition of radish root elongation compared to the control treatment (Fig. 2). This inhibition of growth was equivalent to that obtained when roots were exposed to IAA at a concentration of ca. 5 × 10⁵ M in the root bioassay (Fig. 1). In contrast, the average length of roots grown in the presence of strain 299XYLE, the isogenic ipdC-negative mutant with a greatly reduced ability to synthesize IAA, was not significantly different from that of roots exposed to KP buffer only (Fig. 2). The mean population size of strain 299XYLE on roots was about twofold lower than that of the parental strain (Fig. 2). These results were similar to those obtained in replicate experiments with strain 299XYLE as well as with strain 299MX149, another ipdC-negative mutant described earlier (2).

View larger version (27K): [in this window] [in a new window]   FIG. 1.   Effect of indole derivatives on the elongation of radish roots. Roots were grown in growth packs wetted with tryptophan (), indolepyruvate (), indoleacetaldehyde (), TOL (), or IAA () solutions of increasing concentrations. Distilled water was used for the control treatment (×). Each value is the mean for five replicate packs containing 10 roots each; error bars indicate ±1 standard error of the mean.

View larger version (30K): [in this window] [in a new window]   FIG. 2.   Effect of E. herbicola 299R and 299XYLE on the elongation of radish roots (solid bars) and their respective populations (hatched bars) on 2-cm root tips, as measured in the root bioassay for IAA. The control treatment (CTL) consisted of KP buffer. Values marked by the same letter were not significantly different, as determined by the Duncan's multiple-range test, at a P value of 0.05.

Estimation of population ratios in vitro. Strain 299XYLE grew at the same rate and achieved the same cell concentration as strain 299R in minimal A medium with or without 0.02% L-tryptophan (data not shown). Moreover, the mean ratio of the number of cells of strain 299R to that of strain 299XYLE was approximately 1.0 throughout 23 generations of growth in minimal A medium (3 days) after inoculation with equal proportions of the two strains (Fig. 3). Similarly, the ratio of these two strains did not change significantly during growth in minimal A medium containing 1.28% NaCl (Fig. 3), a culture condition under which cells experience low osmotic pressure and under which the transcriptional activity of ipdC is increased 32-fold (3).

View larger version (21K): [in this window] [in a new window]   FIG. 3.   Cell density dynamics of E. herbicola 299R and 299XYLE in a competition experiment in which both were coinoculated in a ratio of 1:1 in minimal A medium () and minimal A medium containing 1.28% NaCl (). Values represent the mean of the ratio of the cell concentration of strain 299R to that of strain 299XYLE, and error bars represent the standard error of the mean.

Population dynamics of E. herbicola 299R and 299XYLE on beans in the greenhouse. Strain 299XYLE grew at approximately the same rate as strain 299R and achieved population sizes on bean leaves similar to those of strain 299R when inoculated alone (Fig. 4). However, the rate of growth of 299XYLE was slightly lower than that of the parental strain when both were present on the same leaves. This difference in growth rate was apparent as an increase in the ratio of the population size of 299R to that of the mutant in the 10 to 18 h immediately following inoculation (Fig. 5). The ratios of 299R to 299XYLE increased as much as 1.8-, 1.7-, and 2-fold when these strains were applied to the plants in proportions of 1:1, 1:10, and 10:1, respectively. Regression analysis of the ratio against the time during which an increase in the ratio was observed showed a significant linear relationship for all three experiments (P < 0.01). A two-tailed t test performed on the slope of these regressions also provided evidence that the increase in the ratio of the population size of strain 299R to that of strain 299XYLE over time was significant (t_slope, 3.17, 2.82, and 3.35 for inoculation ratios of 1:1, 1:10, and 10:1, respectively; P < 0.01 for all three experiments). However, the relative population sizes of the strains were not maintained after cell growth had slowed (ca. 15 to 18 h after inoculation), and the ratio of 299R to 299XYLE decreased about 50% once the strains had established maximum population sizes on the leaves. Similar trends in ratios were observed in replicate experiments in the greenhouse; the parental strain consistently reached significantly higher population sizes than 299XYLE.

View larger version (18K): [in this window] [in a new window]   FIG. 4.   Population dynamics of E. herbicola 299R () and the isogenic IAA-deficient xylE insertional mutant, 299XYLE (), after inoculation of each strain individually onto bean plants in a greenhouse experiment. The error bars represent ±1 standard error of the mean of the log-transformed bacterial population sizes.

View larger version (14K): [in this window] [in a new window]   FIG. 5.   Change in the ratio of the population of E. herbicola 299R to that of E. herbicola 299XYLE over time after the strains were coinoculated in proportions of 1:1 (A), 1:10 (B), and 10:1 (C) onto bean plants in a greenhouse experiment. Values represent the mean of the ratios of the arithmetic population sizes of the two strains, determined from individual leaves. Error bars represent ±1 standard error of the mean.

Population dynamics of E. herbicola 299R and 299XYLE on pear flowers in the field. The population size of the mutant strain deficient in IAA production was similar to that of parental strain 299R at a given time after both were applied to pear flowers alone, although the mean population sizes of the IAA-deficient mutant were generally smaller than those of parental strain 299R throughout the sampling period during an experiment in 1994 (Fig. 6). It is noteworthy that strain 299R attained and generally maintained slightly larger populations despite an initial cell density on the flowers lower than that of strain 299XYLE (Fig. 6). A similar trend was observed after the strains were coinoculated in a proportion of 1:1. It should be noted that the log value of the ratio of the parental strain population size to that of the IAA-deficient mutant increased from 0.2 1 day after inoculation to greater than 0 throughout most of the sampling period (Fig. 7) (because of the high variability in the value of the individual ratios at any given sampling time in the field experiments, the ratio of the population size of strain 299R to that of strain 299XYLE was log transformed to achieve a normal distribution of the data). Qualitatively similar results were obtained in repeat experiments conducted in the spring of 1995 and the spring of 1996 (data not shown). Regression analysis of the log-transformed ratio of IAA-producing and non-IAA-producing strains from all of the individual field samples against time showed a relationship between ratio and time (P = 0.06) (Fig. 8). The low R² value reflected the high variability of the population ratios observed on individual flowers within a few days after inoculation. Although the population sizes of both strains were initially very similar after coinoculation onto pear flowers, the proportions of the two strains in individual flowers varied greatly, even within a given treatment, after about 4 days (Fig. 7 and 8). This same phenomenon was observed in repeat experiments conducted in 1995 and 1996 (data not shown). Regression analysis also revealed that the parental strain comprised a generally higher proportion of the recoverable population than the IAA-deficient mutant over the sampling period (t_slope = 1.87; P < 0.05) (Fig. 8).

View larger version (27K): [in this window] [in a new window]   FIG. 6.   Population dynamics of E. herbicola 299R () and the isogenic IAA-deficient mutant, 299XYLE (), after inoculation of each strain individually onto pear flowers in the field in 1994. Error bars represent ±1 standard error of the mean of the log-transformed bacterial population sizes. Samples collected before approximately 9 days after inoculation consisted of immature or mature pear flowers and thereafter consisted of immature fruits.

View larger version (32K): [in this window] [in a new window]   FIG. 7.   Population dynamics of E. herbicola 299R () and 299XYLE () and change in the ratio () of their respective populations after coinoculation in a proportion of 1:1 onto pear flowers in the field in 1994. Values represent the mean of the log-transformed population sizes of 299R and 299XYLE and the mean of the log-transformed ratio of the untransformed population sizes of these strains on individual flowers. Error bars represent ±1 standard error of the mean. For ratio data, only the standard error of the mean below the mean is shown for clarity. Numbers in parentheses on the right-hand y axis represent the antilog of the log-transformed ratio of population sizes, for easier interpretation of the data.

View larger version (22K): [in this window] [in a new window]   FIG. 8.   Regression against time of the log-transformed ratios of the arithmetic population sizes of E. herbicola 299R and 299XYLE from individual pear flowers or young fruits after inoculation of the flowers with a 1:1 mixture of these strains for the entire sampling season in 1994. The line drawn represents the linear regression y = 0.006x + 0.03 (P = 0.06, R2 = 0.01).

View larger version (21K): [in this window] [in a new window]   FIG. 9.   Regression of the change in the ratio of the population size of E. herbicola 299R to that of E. herbicola 299XYLE against the change in the combined population sizes of these two strains when coinoculated onto pear flowers for three consecutive field seasons. Each axis represents the slope calculated from the regression of the mean of the log-transformed population ratio (ordinate) and the mean of the log-transformed population size (abscissa) against time for individual periods when the combined populations of the two strains were either generally increasing or generally decreasing. Numbers in parentheses on the ordinate represent the antilog of the rate of change in the mean of the log-transformed ratio per day, for easier interpretation of the data. The line drawn represents the linear regression y = 0.3x  0.0008 (P = 0.003, R2 = 0.85).

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The occurrence of IAA production in tryptophan-supplemented cultures of E. herbicola strains and many other nonpathogenic plant-associated bacteria is common (19). Increased transcriptional activity of ipdC during the growth of E. herbicola 299R on plant surfaces provides some evidence for the bacterial production of IAA in the phyllosphere (3). In this study, we found more direct evidence for the production of IAA on plants by E. herbicola by using a root bioassay for IAA. Because roots are very sensitive to IAA, their elongation is inhibited by relatively low concentrations of exogenous IAA (27). A linear relationship correlating the inhibition of root elongation with the amount of IAA produced by rhizobacteria in cultures was observed previously (20). The differential effects on root elongation of strains 299R and 299XYLE showed that the factor responsible for the inhibition of root growth by strain 299R was the exogenous production of IAA, since these isogenic strains differed only in their ability to synthesize IAA. The observation that the population size of strain 299XYLE on roots was twofold lower than that of strain 299R cannot account solely for its lesser effect on root growth, since a resultant twofold reduction in IAA production from wild-type levels should have resulted in only a negligible reduction in root inhibition by the IAA-deficient mutant (Fig. 1). The root bioassay thus provides further evidence that strain 299R produces IAA on plant surfaces and not solely in culture. This bioassay should be a good surrogate for epiphytic growth, since the chemical environment of roots is probably similar to that of leaves and since many epiphytes (such as strain 299R) grow well in the rhizosphere. These results thus indicate that bacterial IAA synthesis can affect the normal physiology of plant cells. The differential effects of the two strains on root growth also indicate the importance of the ipdC gene in the production of IAA by strain 299R on plants.

The use of coinoculations and estimation of ratios of the populations of coinoculated strains over time allowed the differences in behaviors between isogenic IAA-producing and IAA-deficient bacterial strains on bean plants and pear flowers to be more unambiguously determined than comparisons of the population size of each strain inoculated singly, as has been done in most studies (21, 28, 31). This method provided greater statistical power in population studies performed under natural conditions, in which a changing environment is usually an important source of variation in population sizes, since the estimation of the ratio of two strains sharing a common environment inherently contains an internal control (16). That is, large leaf-to-leaf variations caused by sampling and environmental variations do not overshadow differences in the population sizes of individual strains since, by sharing the same leaf, the two strains exhibit similar leaf-dependent population sizes.

The xylE gene enabled us to monitor each strain in mixed populations on the same isolation plate and thus to accurately compute population ratios. The nearly constant ratio of strain 299R to strain 299XYLE in mixed cultures over 23 generations, even under conditions inducing higher levels of ipdC transcription and therefore of xylE expression, indicated that XylE activity affected neither the ability of strain 299XYLE to grow in cultures, nor its plating efficiency. Wilson and Lindow (32) showed in de Wit replacement experiments that the introduction and expression of xylE within the iceC gene in the P. syringae Cit7 genome, shown not to contribute to epiphytic fitness (18), did not reduce the ability of this bacterial epiphyte to colonize the phyllosphere. On the basis of these observations, it was not expected that the presence of xylE per se would affect the ability of E. herbicola to grow and survive on plant surfaces. Moreover, although pleiotropic effects from insertional mutagenesis of ipdC cannot be ruled out, the ipdC gene is well characterized and the function of its product is very specific (2, 14); pleiotropic effects from its inactivation are therefore unlikely. Additionally, derivatives of strain 299R that contained an insertional mutation in the region within about 1,000 nucleotides downstream of ipdC grew as well as the wild type in minimal A medium (data not shown). Thus, there is little evidence to suggest that the lower growth rate of strain 299XYLE is due to a polar effect from insertional mutagenesis of ipdC.

The greenhouse studies on the comparative behaviors of the parental strain and its IAA-deficient derivative clearly demonstrated a difference in the extent of growth of these two strains when coinoculated onto bean leaves. Independent of the initial ratio of the two strains, the population of the parental strain increased to approximately twice that of the mutant strain during the period of active growth and colonization of bean leaves. This change in ratio, when normalized per number of cell generations that occurred during growth on beans, represented a selection rate constant (16, 17) of 0.067 per cell generation. Using a similar approach of studying changes in the ratio of two competing bacterial populations, Lenski et al. (17) demonstrated the decreased competitive fitness of plasmid-bearing Escherichia coli cells compared to plasmid-free cells and reported a selection rate constant of 0.025 per cell generation. Thus, because of the statistical power of the computation of ratios from mixed populations, the twofold increase in the ratio observed in our greenhouse studies represents a significant enhancement in the competitive fitness of strain 299R, due to its ability to produce IAA, particularly when considered over evolutionary time frames. This change in the proportion of IAA-producing to IAA-deficient strains in mixed populations on leaves appears also to reflect a plant-specific benefit of IAA production, since no difference in the growth of these two strains was noted in culture. This benefit may be mediated by the increased leakage of nutrients from plant cells in the vicinity of IAA-producing bacteria colonizing the plant surface. IAA affects plants at very low concentrations and promotes cell wall loosening during cell elongation (30). Exogenously applied auxin can stimulate the release of large quantities of monosaccharides and oligosaccharides from the plant cell wall (11, 12). As we previously hypothesized (1), a similar release of nutrients from plant cells in response to IAA produced by epiphytic bacteria on plants may confer upon them a selective advantage.

The observation that the growth rate of wild-type strain 299R was consistently higher than that of the IAA-deficient mutant on leaves is noteworthy (Fig. 5). Considerable circumstantial evidence suggests that the nutrient concentration on leaves is low. Chet et al. (4) and Fokkema and Lorbeer (10) reported that the concentrations of glucose and other sugars on wetted leaves were in the range of 3 to 20 mg/liter. Our measurements of sugar concentrations on bean leaves grown under the conditions used here also were about 10 µg per leaf, resulting in about 10 mg/liter if the leaf were fully wetted (22). The growth rate for strain 299R on leaves (0.61 generation/h) (Fig. 4) was only about half that in minimal A medium at the same temperature. Monod as cited by Dabes et al. (6) has shown that the growth rate of the enteric bacterium E. coli, which is closely related to E. herbicola, is strongly dependent on nutrient concentration up to concentrations of about 50 mg/liter; half-maximal growth occurs at a nutrient concentration of about 10 mg/liter. We speculate that the relatively low growth rate of strain 299R and other epiphytes on plants may be due to the low concentration of nutrients found in this habitat. Indeed, measurements of the amount of nutrients leached from plant foliage (4, 22) are consistent with this conjecture. If we assume, based on this circumstantial evidence, that bacterial cells on leaves experience low nutrient concentrations, then we would expect from the work of Monod, as cited by Dabes et al. (6), that the growth rate on leaves would vary nearly directly with the nutrient concentration. The generation time for strain 299R was observed to be only 0.93 that of the IAA-deficient mutant on bean plants (Fig. 5). This finding suggests that the nutrient concentration in the vicinity of the parental strain was about 7% higher than that near the mutant strain. If the nutrient concentrations on leaves were as low as 10 mg/liter, as we suggested above, only about an additional 1 mg of nutrients per liter would need to be made available to the IAA-producing strain to achieve the higher growth rate reported in our study.

The similar increase in the ratio observed when the initial proportion of strain 299R was 10-fold larger than that of 299XYLE, compared to when it was present in an equal or smaller proportion, indicates that the parental strain did not contribute to the epiphytic fitness of the mutant strain under the tested conditions. The failure of the parental strain to complement the mutant strain on plant surfaces suggests that the release of bacterial IAA has a very localized effect on the leaf environment and that its benefit may be limited to the bacterial cells that produce IAA.

Very similar growth-dependent changes in the ratio of the IAA-producing strain to the IAA-deficient strain were observed in greenhouse studies of bean leaves as well as in several field studies of pear flowers. Two distinct trends in the ratio of the population sizes of the two strains were detected in both studies: (i) a general increase in the proportion of the IAA-producing strain coincident with an increase in the combined populations of the two strains and (ii) a general decline in the proportion of the IAA-producing strain when total population sizes decreased. The first trend indicates that the production of IAA conferred a selective advantage to the parental strain during periods of active colonization of the phyllosphere. It is noteworthy that the induction of ipdC occurs during the active phase of growth of strain 299R on plants (3), indicating that IAA is produced during this period. This finding supports our conjecture that IAA synthesis contributes to the growth of strain 299R on plant surfaces. Although the production of IAA conferred some advantage to E. herbicola in the colonization of the phyllosphere, it was not beneficial for survival in this habitat, since the increased ratio of the population of strain 299R to that of strain 299XYLE was not maintained upon the cessation of growth on bean leaves and pear flowers. It is likely that the increased population size of parental strain 299R attained in the presence of the production of IAA during multiplication in the phyllosphere would not have been sustained if, for example, IAA-induced nutrient leakage decreased due to a cessation of bacterial IAA production and substrate availability subsequently became low. In this situation, the ratio of the IAA-producing strain to the IAA-deficient strain would have declined. These two trends in population size and ratio changes were detected consistently not only in our greenhouse studies but also in experiments with pear flowers during three consecutive field seasons. Indeed, the correlation between the dynamics of bacterial population sizes and the dynamics of population ratios that was demonstrated by regression analysis of the combined data from all three field studies was quite strong. This result clearly indicates that a benefit of IAA production occurs primarily when cells can exploit resources in the phyllosphere for further growth.

The results from the greenhouse and field studies revealed that the relative abundance of strain 299R increased about twofold compared to that of strain 299XYLE over a period of time when net population size changes indicated that at least 9 cell generations had occurred. Although IAA production did not confer upon strain 299R large qualitative differences in its ability to exploit plant surfaces compared to that of the isogenic IAA-deficient mutant strain, the apparent benefit to epiphytic fitness conferred by IAA biosynthesis is large when considered over evolutionary time periods. The overall fitness of a bacterial strain is most likely contributed by many individual traits, and large differences in the fitness of nearly isogenic strains are unlikely. Indeed, many studies of microbial evolution have shown only small differences in the ratio of two nearly isogenic strains grown together in the same environment over thousands of cell generations; these small differences are nonetheless considered very significant in terms of their impact on bacterial competitive fitness (8, 24). Thus, the common occurrence of IAA production among strains of E. herbicola may be explained by the fitness benefits that it confers during the extensive epiphytic life of this bacterium. These benefits may be related to plant-mediated changes in resource availability to E. herbicola on plant surfaces.