Impact of the Microscale Distribution of a Pseudomonas Strain Introduced into Soil on Potential Contacts with Indigenous Bacteria

Abstract

Soil bioaugmentation is a promising approach in soil bioremediation and agriculture. Nevertheless, our knowledge of the fate and activity of introduced bacteria in soil and thus of their impact on the soil environment is still limited. The microscale spatial distribution of introduced bacteria has rarely been studied, although it determines the encounter probability between introduced cells and any components of the soil ecosystem and thus plays a role in the ecology of introduced bacteria. For example, conjugal gene transfer from introduced bacteria to indigenous bacteria requires cell-to-cell contact, the probability of which depends on their spatial distribution. To quantitatively characterize the microscale distribution of an introduced bacterial population and its dynamics, a gfp-tagged derivative of Pseudomonas putida KT2440 was introduced by percolation in repacked soil columns. Initially, the introduced population was less widely spread at the microscale level than two model indigenous functional communities: the 2,4-dichlorophenoxyacetic acid degraders and the nitrifiers (each at 10⁶ CFU g⁻¹ soil). When the soil was percolated with a substrate metabolizable by P. putida or incubated for 1 month, the microscale distribution of introduced bacteria was modified towards a more widely dispersed distribution. The quantitative data indicate that the microscale spatial distribution of an introduced strain may strongly limit its contacts with the members of an indigenous bacterial community. This could constitute an explanation to the low number of indigenous transconjugants found most of time when a plasmid-donor strain is introduced into soil.

Soil bacteria are capable of performing a wide range of activities, some of which result in plant growth promotion or in soil pollutant mineralization. Bioaugmentation of soil, which is the addition of microorganisms to enhance a specific biological activity (53), has been proposed to take advantage of these bacterial abilities. The beneficial roles of the bacterial inoculants are performed either directly by the introduced bacteria (51) or by the means of plasmid-mediated bioaugmentation (48). The latter consists of the dissemination among well-established soil indigenous communities of one or several genes carried by the introduced strain. It is crucial that introduced bacteria reach their target, whether that is a pollutant to degrade, a plant root to colonize, or the indigenous bacteria targeted by the plasmid-mediated bioaugmentation (51). As bacteria occupy only a small proportion of the soil poral space (35), the contacts between the introduced bacteria and their target may be limited and are dependent upon their spatial distribution. It has been proposed that an inadequate location of introduced bacteria within the soil fabric could hamper contact between bacteria and pollutants and could explain the persistence of these pollutants and thus the failure of inoculation for bioremediation purposes (42, 53). However, experimental evidence supporting this hypothesis is rare (13, 45). Similarly, plasmid-mediated bioaugmentation may be hampered by an inadequate location of introduced bacteria. Several microscopic observations suggested that an introduced strain would rather occupy the large pores, while indigenous bacteria can be found embedded in organic matter or clay (9, 36). Recorbet et al. (40), who compared the location of introduced and indigenous bacteria with respect to soil fractions, also concluded that whereas indigenous bacteria were mainly associated with the inner part of soil aggregates, only a minority of introduced Escherichia coli were found within or on this soil fraction.

The need of adequate contact between introduced bacteria and their target has been recognized in the literature but has usually not been considered at the bacterial scale. Most of the time, the emphasis has been placed on the transport of inoculated cells through the soil profile (11, 19, 46). Nevertheless, the necessity of microscale approaches to improve our understanding of the fate and activity of microbial inoculants has recently been emphasized, especially considering the high microscale spatial heterogeneity of the soil ecosystems (44, 51, 52). To the best of our knowledge, the microscale spatial distribution of introduced bacteria has been neither quantitatively characterized nor compared to the distribution of indigenous bacteria. Recently, there has been a renewed interest in the microscale spatial distribution of indigenous bacteria in soil. Using microscopic techniques, Nunan and coworkers have described the in situ distribution of soil bacteria at multiple scales (31) and with respect to soil porosity (32). The three-dimensional microscale distributions of specific autochthonous functional communities have been characterized as well, by an approach based on soil microsampling for with nitrifiers (14) and 2,4-dichlorophenoxyacetic acid (2,4-D) degraders (34).

In this work, we characterized the microscale spatial distribution of a gfp-tagged derivative of Pseudomonas putida KT2440 introduced into soil columns by percolation. KT2440 was chosen as a model because it is considered a good candidate for deliberate release in agroecosystems (29, 38, 47). Subsequent to inoculation, the columns were submitted to treatments simulating conditions possibly encountered in the natural environment, including water or substrate percolation and static incubation of the soil for 6 weeks. The characteristics of the spatial distribution of introduced bacteria (size and number of colonized soil patches) were determined at a 125-μm spatial resolution by using a specific sampling strategy combined with statistical data analysis (6). The spatial distribution of the introduced bacteria was then statistically compared to the distribution of two indigenous bacterial communities (nitrifiers and 2,4-D degraders) for which the same spatial characteristics were available, and the probabilities of cooccurrence of the indigenous and introduced bacteria in the same soil microsamples were evaluated.

MATERIALS AND METHODS

Soil and soil columns.

The soil was a sandy clay loam (Alfisol, typic hapludalf) sampled in a cultivated corn field (La Côte Saint André, Isère, France). The top 30 cm of the soil was sampled, sieved (≤4 mm), and kept at 15°C and 18% gravimetric water content. The soil characteristics were as follows: organic matter content, 2.5%; C/N molar ratio, 11; pH 7.0; water-holding capacity, 25.8 g H₂O per 100 g dry soil; dry bulk density, 1.34 g cm⁻³ (20, 39).

Soil columns consisted of a 5-cm-inside-diameter, 12-cm-high acrylic plastic cylinder. The columns were prepared by sequentially introducing 80-g portions of wet soil, each packed in the same manner to reach homogenous dry bulk density (1.46 ± 0.02 g cm⁻³) and porosity (45% ± 1%). All experiments were performed at 22°C.

Bacterial strain and culture media.

The construction of P. putida KT2440 harboring the P_A1/04/03-RBSII-gfpmut3*-T0-T1 transposon was performed as described by Heydorn et al. (17). This transposon cassette carried by pJBA28 (2) and was inserted at random positions into the chromosome of the recipient P. putida KT2440 by triparental mating. Filters carrying a mixture of donor, recipient, and helper strains (E. coli HB101 carrying plasmid RK600) (2) in a ratio of 1: 3: 1 were incubated overnight at 30°C on LB plates (27). The cells were washed and then suspended in 0.8% NaCl, and serial dilutions were plated on selective medium. The selected gfp-tagged strain, FBC004, showed a stable green fluorescent phenotype. The gfp marker constitutes an insignificant metabolic burden to KT2440 (24). FBC004 was routinely grown in LB medium. M9 minimum medium (27) with 1.22 g liter⁻¹ of benzoate as the sole C source, supplemented with cycloheximide (200 μg μl⁻¹) and kanamycin (25 μg μl⁻¹), was used to selectively cultivate FBC004 from soil suspensions. To detect the presence of FBC004 in soil microsamples, the microsamples were incubated in 1.5 ml of M9 benzoate medium in 24-well microplates at 30°C. In the presence of FBC004, the medium turned green after 2 or 3 days of incubation. The presence of the gfp gene was verified using a multilabel microplate reader (Victor, Perkin-Elmer, Courtaboeuf, France) set at 485 and 510 nm (excitation and emission wavelengths, respectively). To enumerate FBC004 in soil microsamples, M9 benzoate medium plates supplemented with kanamycin (25 μg ml⁻¹) were used. On these plates, very few bacteria other than FBC004 formed colonies, and they did not present the typical green shine of the FBC004 colonies. When the gfp phenotype of a colony was doubtful, the colony was grown in liquid medium and checked as indicated above.

For FBC004 introduction into soil columns, FBC004 was grown overnight in liquid LB at 28°C. The culture was then centrifuged (5000 × g; 15 min). The supernatant was discarded, and the cells were suspended in KCl (1.16 g liter⁻¹, equivalent to the ionic strength of the soil solution), to obtain an optical density at 600 nm (OD₆₀₀) of 0.003, 0.03, 0.1, or 1.

Column experiments.

A total of 23 soil columns were prepared, inoculated, and sampled after various incubation times. Due to the destructive sampling, the dynamics of FBC004 microscale spatial distribution and abundance after introduction had to be studied on different columns. The column names and corresponding cell densities used for the introduction of FBC004 are listed in Table 1. Replicates of the same experiment were carried out and are referred to as A, B, and C. They sometimes differed by their inoculum densities (Table 1).

TABLE 1.

Abundance of introduced FBC004 and characteristics of the theoretical spatial distributions compatible with the experimental sampling results for the 23 columns used in this study

Column name	OD of cell suspension used for FBC004 percolation	Mean FBC004 abundance (CFU g−1 dry soil)	Characteristic(s) of compatible theoretical spatial distributions
			Diam of colonized patches (μm)	No. of colonized patches g−1
T0	1	1.6 × 107	≤205	560-935
KClA.T0	1	1.1 × 108	≤490	285-1,654
KClA.T7	1	2.3 × 106	≤325	320-995
KClB.T0	0.1	3.1 × 106	≤325	630-1,600
KClB.T7	0.1	1.5 × 105	≤540	360-2,095
BenzA.T0	0.1	5.1 × 106	≤205	810-3,090
BenzA.T7	0.1	1.3 × 106	325-415	1,705-4,500
BenzB.T0	0.03	7.0 × 104	≤110	1,930-2,645
BenzB.T7	0.03	3.0 × 106	205-415	3,105-15,695
BenzC.T0	0.003	2.6 × 103	110-325	580-1,930
BenzC.T7	0.003	1.4 × 105	205	2,920-3,535
IncA.T0	0.1	2.8 × 106	205	1,190-3,035
IncA.T7	0.1	4.0 × 106	≤110	5,070-6,175
IncA.T14	0.1	3.1 × 106	205	3,615-15,990
IncA.T28	0.1	7.3 × 105	205	3,695-27,290
IncB.T0	0.1	3.0 × 106	NCTDa	NCTD
IncB.T14	0.1	1.9 × 105	NCTD	NCTD
IncB.T30	0.1	1.1 × 105	NCTD	NCTD
IncB.T42	0.1	2.4 × 105	205, 755	1,090-13,200
IncC.T0	0.1	8.2 × 106	NCTD	NCTD
IncC.T14	0.1	1.1 × 105	205	5,725-5,855
IncC.T31	0.1	3.1 × 105	205	8,410-47,415
IncC.T45	0.1	8.0 × 105	325-755	1,505-17,245

^aNo compatible theoretical spatial distribution was found.

A suspension of FBC004 cells, prepared as described in the previous section and continuously stirred, was applied to the top of the soil columns at a flow rate of 3 ml h⁻¹ for 12 h. At the end of cell introduction, the soil water content was 24.1% ± 1.6%. Nine columns, designated T0, were sampled at the end of the 12-h FBC004 percolation and were used as controls in the study of FBC004 spatial dynamics.

After FBC004 introduction, the other columns were either percolated for 7 days with a KCl solution (1.16 g liter⁻¹) (columns KCl.T7), or percolated for 7 days with a benzoate solution (0.366 g liter⁻¹, supplemented with 1.13 g liter⁻¹ KCl and pH adjusted to 6 with KOH) (columns Benz.T7), or incubated without continuous percolation (columns Inc). In this last case, 3 ml of KCl aqueous solution was applied on the top of the columns every 2 or 3 days to prevent soil desiccation. Inc columns were sampled after 0, 1, 2, and 4 weeks (IncA) or every 2 weeks (IncB and IncC).

Soil sampling and sample processing.

The top 1.5 cm of each soil column was discarded because it had been submitted to the impact of the solution drops and was considered nonrepresentative of the entire soil column (33). Similarly, the soil in contact with the column wall was discarded because preferential flow may have occurred in this zone.

A sample of approximately 8 cm³ was taken at the 1.5-cm depth in the center of the column to study the microscale spatial distribution of the introduced bacteria, following the method of Grundmann et al. (14). This soil sample was randomly dissected into microsamples with a sterile scalpel. The size of the microsamples was determined on a grid gauge; about 100 microsamples with diameters measuring approximately 125, 250, 500, 1,000, and 1,500 μm were selected and individually transferred to the M9 benzoate liquid medium. The microsamples were referred to as size 125, 250, 500, 1000, and 1500, depending on their diameter. No samples below size 125 were dissected because sample desiccation could have occurred quickly, impairing the survival of the FBC004 cells the samples may have contained. For each microsample size, the observed percentage of microsamples harboring FBC004 was recorded. This is referred to as presence frequency. Additionally, 35 size-500 microsamples and 35 size-1000 microsamples were individually thoroughly dispersed in 500 μl of NaCl (8 g liter⁻¹) and spread on M9 benzoate agar plates for FBC004 enumeration.

The FBC004 strain was enumerated in triplicate by the spread plate technique on 5-g soil samples. A soil fractionation was also performed in triplicate on 5-g soil samples: the inner and outer compartments of the soil sensu Hattori (16) were separated by the method of Ranjard et al. (39). This method consists of 15 successive washings of each soil sample in 50 ml sterile NaCl (8 g liter⁻¹), using a 250-ml Erlenmeyer flask placed on a rotary shaker. The outer compartment corresponds to the material released in the supernatants. FBC004 strain was enumerated in the two soil fractions.

Spatial analysis.

The spatial distribution of FBC004 was characterized by the method of Dechesne et al. (6). The rationale of this method is to compare experimental FBC004 presence frequencies with the results expected from the sampling of numerous theoretical spatial distributions of known characteristics. Soil is modeled as a three-dimensional grid, the elementary unit of which usually corresponds to the smallest microsample volume. The theoretical distributions are either random distribution of positive (here, colonized by FBC004) elementary units or random distribution of clusters of positive elementary units. These clusters correspond to colonized patches in the soil. The spatial descriptors that define theoretical spatial distributions are the diameter of colonized patches and the number of colonized patches per gram of soil. When a theoretical distribution is compatible with the experimental data, it is accepted as representative of the true distribution with a type I error that is ≤5%. Usually, several theoretical distributions are accepted for a single presence frequency data set. Therefore, a range of spatial descriptor values is usually obtained for an experimental data set. If two experimental data sets analyzed by this spatial analysis method do not share any common accepted theoretical distribution, then the null hypothesis (the spatial distributions corresponding to the two data sets are identical) can be rejected with a type I error that is ≤2.5%. It has to be noted that this method only takes cultivable bacteria into account and that the bacterial distributions are not studied at the scale of the bacterial cells but with a spatial resolution corresponding to the elementary unit (here, a “cube” of soil with sides measuring either 125 μm or 50 μm).

Comparison with the spatial distributions of indigenous bacteria and cooccurrence estimation.

The data obtained for introduced FBC004 were compared to results obtained for indigenous bacterial communities in the same soil and by the same sampling strategy. These communities were the 2,4-D degraders (34) and the NO₂⁻ and NH₄⁺ oxidizers (14; unpublished results). Whereas in the present study and in the study by Pallud et al. (34) the smallest sample size was 125 μm, an additional analysis at a 50-μm resolution was chosen for studying the microscale cooccurrence of introduced FBC004 and indigenous bacterial types. This allowed getting closer to the bacterial microhabitat scale where cell-to-cell contact actually takes place. The presence probabilities in size-50 microsamples of introduced and indigenous bacteria were estimated by performing the spatial analysis of Dechesne et al. (6) (see “Spatial analysis,” above) with a size-50 microsample as the elementary unit. It should be noted that analyzing data sets at a spatial resolution smaller than the smallest microsample actually sampled increases the imprecision of the estimated probability of bacterial presence. Under the hypothesis of the independence of the presence or absence of introduced and indigenous bacteria, the cooccurrence probability is the product of the two presence probabilities. The number of cooccurrences per gram of soil was calculated by multiplying the cooccurrence probability by the number of size-50 microsamples in a gram of soil. The latter parameter was estimated to be 8.6 × 10⁶, considering the dry bulk density of the column (1.46 g cm⁻³) and taking into account the macroporosity between microsamples into account (microsample density, 1.8 g cm⁻³) (6).

RESULTS

FBC004 microscale distribution after introduction into soil.

The presence frequencies of bacteria (FBC004 or members of the model indigenous communities) in microsamples as a function of microsample size are presented in Fig. 1. The curves associated with FBC004 presented similar patterns, except for IncB.T0 and IncC.T0, which presented higher presence frequencies for the size-500 microsamples. For the smallest microsamples (sizes 125 and 250), the presence frequencies of the members of the indigenous communities were consistently higher than those of the introduced FBC004. This indicated that the indigenous bacterial communities were more widely spread in the soil than the introduced bacteria. Interestingly, no link could be found between presence frequencies of FBC004 and its abundance in soil, which depended upon the OD of the cell suspension used for introduction (Table 1). For example, curves associated with KClA.T0 and BenzC.T0 were very similar, in spite a difference of a 5 orders of magnitude between the corresponding FBC004 abundances.

FIG. 1.

Bacterial presence frequency in soil microsamples as a function of microsample diameter for 18 soil columns. Experimental data corresponding to indigenous communities are figured in dashed lines: 2,4-D degraders (34), Nitrobacter like and Nitrosomonas like (14; unpublished results). Continuous lines show FBC004 presence frequencies measured in soil columns sampled after a 12-h percolation with FBC004 cell suspensions. Mean abundances of indigenous or introduced bacteria g⁻¹ soil are indicated in parentheses in the legend box.

The spatial analysis developed by Dechesne et al. (6) was used to determine the spatial characteristics of the FBC004 distribution associated with each FBC004 introduction experiment (Table 1). For the two atypical curves in Fig. 1 (IncB.T0 and IncC.T0), no compatible distributions were found. In all other cases, whatever the abundance of FBC004, random distributions of colonized size-125 microsamples were compatible with experimental presence frequencies. In addition, for four of the replicates, aggregated theoretical distributions were also compatible with experimental results (Table 1): the corresponding diameters of colonized patches were <325 μm, except for one column (≤490 μm for KClA.T0). After FBC004 introduction, the size-125 microsamples colonized by FBC004 were thus either randomly distributed in soil or clustered in “small” patches.

The abundance of FBC004 in size-500 and size-1000 microsamples showed a high variability (Table 2). Microsamples devoid of introduced bacteria, as well as microsamples harboring several thousands of introduced bacteria, coexisted in the columns.

TABLE 2.

Frequency distribution of size-500 microsamples into seven FBC004 abundance classes for the column BenzA.T0^a

Parameter	Abundance class (no. of FBC004 CFU/size-500 microsample)
	0	1-9	10-99	100-199	200-299	300-399	>400
Frequency of abundance classes (%)	64.1	5.7	3.8	0.0	1.9	1.9	22.6

^aFifty-three microsamples were sampled immediately after FBC004 introduction.

In each column, FBC004 counts in the soil fractions indicated that the abundance of introduced bacteria were 2 orders of magnitude higher in the outer fraction than in the inner fraction (data not shown).

FBC004 microscale distribution after a 1-week KCl percolation.

During KCl percolation, the abundance of introduced bacteria significantly (P < 0.05) decreased from 1.1 × 10⁸ to 2.3 × 10⁶ CFU g⁻¹ and from 3.1 × 10⁶ to 1.5 × 10⁵ CFU g⁻¹ for experiments KClA and KClB, respectively. Accordingly, the average number of FBC004 per microsample decreased as indicated by a significant decrease in the frequency of size-500 microsamples harboring >400 FBC004 cells (P = 0.0011 and P < 0.001 for KClA and KClB, respectively) (data not shown). Nevertheless, no clear change in the presence frequencies was observed (Fig. 2A), and no spatial change was detected at the spatial resolution used (125 μm) (Table 1).

FIG. 2.

Influence on FBC004 presence frequencies of a 1-week KCl percolation (A) and a 1-week benzoate percolation (B). The two graphs show results obtained in replicate columns: KClA and KClB (A) and BenzB and BenzC (B), differing only by their initial FBC004 inoculum (Table 1). “T0” indicates that the columns were sampled at the end of FBC004 introduction and “T7” indicates that sampling occurred after a 7-day percolation.

The abundance of FBC004 remained 2 orders of magnitude higher in the outer compartment than in the inner compartment (data not shown).

FBC004 microscale distribution after a 1-week benzoate percolation.

After benzoate percolation, the FBC004 abundance per gram of soil either increased or remained the same (Table 1). It significantly increased (P < 0.05) when the initial FBC004 abundance was low (7 × 10⁴ and 2.6 × 10³ CFU g⁻¹ for experiments BenzB and BenzC, respectively), while no increase in abundance was observed for BenzA, where the initial FBC004 abundance was high (5 × 10⁶ CFU g⁻¹). FBC004 cells remained 2 orders of magnitude more abundant in the outer than in the inner soil fraction (data not shown).

In all cases, the benzoate percolation provoked an increase in FBC004 presence frequencies (Fig. 2B), indicating that FBC004 cells had become more widely distributed in the soil. The spatial analysis confirmed that significant (P < 0.025) spatial changes occurred during the percolation (Table 1). Increases in the number of colonized patches per gram of soil and in the patch diameter contributed to this spatial spreading. Indeed, a significant (P < 0.025) increase in the number of colonized patches was shown for BenzB and BenzC, and a significant (P < 0.025) increase in colonized patch diameter was shown for BenzA and BenzB (Table 1).

Contrary to what was observed after KCl percolation, the microsamples densely colonized by FBC004 were maintained during the benzoate percolation, since the frequencies of the microsamples colonized by >400 FBC004 cells did not significantly decrease during benzoate percolation (data not shown).

FBC004 microscale distribution during long-term incubation.

The FBC004 abundance significantly decreased (P < 0.05) from 10⁶ to 10⁷ CFU g⁻¹ after cell introduction to 10⁵ to 10⁶ CFU g⁻¹ at the end of incubation (Table 1). The enumeration of FBC004 in soil fractions was only performed for experiment IncA, and no significant increase in FBC004 abundance was detected in the inner fraction (data not shown).

The presence frequencies increased with time for the three replicates (Fig. 3A to C), indicating a spatial spreading of the introduced bacteria. In the case of the experiment IncC, the absence of spatial spreading during the first month may have been due to the drop below 20% of the soil water content. A rewetting of the soil, performed between days 30 and 42, seemed to restore the spreading process. The spatial distributions at the end of incubation differed for the three replicates: for experiment IncA, small colonized patches (diameter, ≤205 μm) were identified, while larger ones were identified for experiment IncC (diameter, 325 to 755 μm) (Table 1). Furthermore, for several sampling dates the spatial analysis failed to identify theoretical distributions compatible with experimental results, suggesting that the real bacterial distributions were more complex than those considered in the spatial analysis. Together with the decrease in the frequencies of noncolonized microsamples (Fig. 3), there was an increase in the frequencies of the microsamples harboring 1 to 200 and 100 to 300 FBC004 cells for the size-500 and size-1000 microsamples, respectively (Fig. 4A and B).

FIG. 3.

Modification with time of FBC004 presence frequency as a function of soil microsample size for three incubation experiments without continuous percolation: IncA (A), IncB (B), and IncC (C). The number included in the names shown in the insets (e.g., 28 in IncA.T28) indicates the incubation time (in days) that separated FBC004 introduction from the sampling.

FIG. 4.

Modification over time of the abundance of FBC004 in size-500 (A) or size-1000 (B) microsamples for the incubation experiment without continuous percolation IncA. The number included in the names shown in the insets (e.g., 28 in IncA.T28) indicates the incubation time (in days) that separated FBC004 introduction from the sampling.

Cooccurrence estimation.

Estimates of cooccurrence probabilities between introduced and indigenous bacteria in size-50 microsamples were derived, using the spatial descriptors provided by the spatial analysis. The estimates of presence probability in size-50 microsamples ranged from 1.4 × 10⁻⁴ to 3.9 × 10⁻² for FBC004 just after their introduction, whatever their abundance, while for the indigenous 2,4-D degraders at an abundance of 10⁶ CFU g⁻¹, the estimates of presence probabilities ranged from 0.42 to 0.9. Under the hypothesis that the presence or absence of these bacterial types are independent from each other, the expected number of cooccurrences in size-50 microsamples per gram of dry soil would range from 5.08 × 10² to 3.0 × 10⁵ (Table 3). Prior to 2,4-D application, indigenous 2,4-D degraders were characterized by a low abundance (10² CFU g⁻¹) and very low presence frequencies in soil microsamples (34). The cooccurrence of introduced FBC004 with a member of this community would then be rare (Table 3). Benzoate percolation or a 1.5-month incubation resulted in a spatial spreading of FBC004 cells. As a result, the expected number of cooccurrences between introduced and indigenous bacteria was high in these cases, compared to situations immediately after FBC004 introduction.

TABLE 3.

Estimated ranges of cooccurrence probability and of cooccurrence numbers per gram of dry soil for introduced FBC004 and indigenous 2,4-D degraders^a

Introduced/indigenous bacterial combination	Probability of cooccurrence in a size-50 soil microsample	No. of cooccurrences g soil−1
FBC004 (T0) and 2,4-D degraders (102)	6.7 × 10−9-3.8 × 10−4	5.8 × 10−2-3.2 × 103
FBC004 (T0) and 2,4-D degraders (106)	5.9 × 10−5-3.5 × 10−2	5.1 × 102-3.0 × 105
FBC004 (Benz.T7) and 2,4-D degraders (106)	2.8 × 10−3-1.8 × 10−1	2.4 × 104-1.5 × 106
FBC004 (Inc.T45) and 2,4-D degraders (106)	1.0 × 10−2-5.2 × 10−1	8.7 × 104-4.5 × 106

^aFour combinations of experiments have been considered to estimate the cooccurrences between FBC004 (T0, immediately after introduction; Benz.T7, after 1-week benzoate percolation; Inc.T45, after 1.5-month incubation); and the indigenous 2,4-D degraders (before or after 2,4-D percolation corresponding to approximately 10² and 10⁶ degraders CFU g soil⁻¹, respectively).

DISCUSSION

The introduction of bacteria bearing catabolic plasmids into soil has been proposed as a way to spread catabolic genes among indigenous soil bacterial communities to favor soil pollutant mineralization (48). The models of conjugal gene transfer in terrestrial environments published so far (4, 22) were based on the mass action model, which is better adapted to well-mixed homogenous bacterial suspensions than to spatially structured environments (23). Because soil is structurally complex, the distribution of introduced bacteria may limit the gene transfers between introduced and indigenous soil bacteria. In the present study, the microscale spatial distribution of P. putida FBC004, a gfp-tagged derivative of P. putida KT2440, was investigated following its introduction into soil by percolation. The distribution of FBC004 immediately after its introduction was less dispersed than the distribution of two model indigenous communities at a 10⁶ CFU g⁻¹ population level. Cell dispersion within the soil fabric was not obtained by increasing the density of cell suspension applied to columns, but a spatial spreading of FBC004 was observed after 1-week benzoate percolations and after 1-month incubations.

Immediately after its introduction, FBC004 distribution was characterized by a high microscale spatial variability. Indeed, within a few cubic centimeters of soil, zones devoid of introduced bacteria neighbored small colonized patches (diameter, <325 μm) (Table 1). The colonized zones often harbored large numbers of FBC004 cells (Table 2). This was particularly true when high inoculum densities were used: for the columns T0 and KClA.T0, the division of the FBC004 abundance obtained using 5-g samples by the number of colonized size-125 soil volumes per gram of soil indicated that the average FBC004 abundance per colonized size-125 volume of soil was >4,000 CFU.

FBC004 spatial distribution right after its introduction corresponds to the pore space immediately accessible to the introduced bacteria. This space depends upon the soil water content (37), the pore network (especially the pore neck size distribution (19), and the volume of the introduced bacterium (11). Because the soil water content was high (18% and 24.5%, prior to and after FBC004 introduction, respectively) but consistent with field values, the smallest micropores were water filled and therefore not readily available to the introduced bacteria (37, 54). During the cell introduction, the water flow was probably limited to the largest pores. Therefore, immediately after their introduction, the FBC004 cells were located in the macropores, as confirmed by their high abundance in the outer soil fraction, which corresponds mainly to the interaggregate porosity (16). This space quickly available to FBC004 was the same whatever the number of introduced cells. That is why increasing the number of FBC004 cells introduced in soil did not result in a significantly larger soil occupation by FBC004. The spatial analysis identified theoretical spatial distributions compatible with experimental FBC004 presence frequencies in soil microsamples. At the 125-μm spatial resolution, the spatial distribution of FBC004 immediately after introduction was characterized as random distributions of “small” colonized patches (diameter, ≤325 μm). Nevertheless, these theoretical distributions have to be interpreted cautiously. Indeed, the macroporosity accessible to FBC004 would probably be partly interconnected rather than randomly distributed.

When soil columns inoculated with bacteria are submitted to a water flow, introduced bacteria may be either transported in the water flow or retained in the columns by adsorption or mechanical filtration (11, 46). FBC004 cells were detected in the effluent of the columns in the first days of KCl percolation (data not shown), indicating that passive movements with water flow resulted in the downward transport of FBC004. Interestingly, this transport involved only a fraction of the introduced cells: while KCl percolation did not provoke a decrease in FBC004 presence frequency (Fig. 2A), a decrease in the number of densely colonized microsamples was observed. This suggests that water flow mainly provoked the flush of fragments of the large cell clusters identified immediately after introduction (Table 2). In spite of these passive cell movements, no significant spatial change at the 125-μm scale was observed after a 1-week KCl percolation. This suggests that the water flow did not result in a significant colonization of new microhabitats by the introduced bacteria in the upper part of the soil columns. The passive bacterial movements with water flow were probably mostly limited to the large pores, which had already been accessible to FBC004. A similar absence in detectable microscale spatial changes due to water flow has also been acknowledged for indigenous 2,4-D degraders at low population levels (10² CFU g⁻¹) (34).

Passive FBC004 movements by diffusion may nevertheless have contributed to the spatial spreading observed after benzoate percolation and long-term incubation. Bacterial growth occurred during the benzoate percolation, and it is known that actively growing soil bacteria are less firmly attached to soil particles than are the remainder of the bacterial community (18), probably because of their cell surface properties (1, 28), and could thus disperse more easily than nondividing cells. In the case of long-term incubations, another mechanism may explain spatial spreading by passive movement: in soil, a reduction in the length of introduced P. fluorescens cells has been observed, probably due to the stress caused by nutrient scarcity (50). This cell shrinkage, which was also detected for a nutrient-starved derivative of P. putida KT2440 (12), may allow the introduced bacteria to access small pores inaccessible to bacteria grown on nutrient-rich medium.

As P. putida FBC004 is motile and the soil water content of the experiments was favorable to motility, active movements probably played a role in the colonization of new microhabitats. Benzoate percolation would have provided an energy source to promote flagella rotation and detachment/attachment from soil surfaces. A similar case of bacterial spatial spreading triggered by a substrate percolation has been described for indigenous 2,4-D degraders (34). P. putida is known to be chemotactically attracted to benzoate (15). It is thus possible that FBC004 chemotactic movements toward pores where benzoate concentration was high (pores not colonized by FBC004) contributed to FBC004 spreading. In the case of benzoate percolation, a marked spatial spreading of the introduced bacteria occurred during the first week, whereas this process was slower for the incubation without continuous percolation. The soil, in spite of its oligotrophy, may nonetheless harbor sufficient amounts of nutrients to support the motility of FBC004. Under this hypothesis, a spatial spreading would also occur for the KCl-percolated columns, although such an event was not detected, probably because of the short period (1 week) separating cell introduction from column sampling. Interestingly, at the end of the long-term incubations, numerous size-500 microsamples harbored between 1 and 300 introduced bacteria (e.g., 85% of the microsamples for IncA.T28) while there were very few immediately after the introduction of FBC004 (13% of the microsamples for IncA.T0) (Fig. 4A). This range of FBC004 abundance per microsample was thus favored during incubation: it may correspond to the carrying capacity for the introduced strain at this scale. At the 5-g soil scale, FBC004 abundance remained high during the 1.5-month incubation (ranging from 10⁵ to 10⁶ CFU g soil⁻¹), in agreement with good survival rates in soil of its parental strain, P. putida KT2440 (29, 38). One can hypothesize that the spatial dynamics of FBC004 is related to its good survival ability: by their movement in soil, FBC004 cells would access new microhabitats and thus new nutrient sources, ensuring the maintenance of high levels of FBC004 abundance at the macroscale. This hypothesis is supported by a study on the role of Pseudomonas fluorescens motility in soil (49). The authors reported that the wild type survived better than its nonmotile derivative, whereas their vertical spreads were similar; the authors suggested that motility enabled the colonization of protective microniches. Nevertheless, another study indicated that the nonmotile derivative of P. fluorescens ANP15 and the wild type survived equally well in soil (3).

Remarkably, the spatial spreading of FBC004 was not associated with a significant colonization of the soil inner fraction, since FBC004 remained mainly in the outer fraction. The low abundance of introduced bacteria in microporosity is a potential problem for plasmid-mediated bioaugmentation, as indigenous soil bacteria are mainly located in the microporosity (9, 21, 26, 32, 36, 39, 40). This low colonization of the inner compartment was unexpected, because this soil is loosely structured. A similar lack of appreciable migration between soil fractions during long-term incubation was reported for Rhizobium leguminosarum introduced into soil (37). The use of soil columns could explain the failure of introduced FBC004 to colonize the soil inner fraction, due to (i) the lack of drying-rewetting cycles which could make the microporosity accessible to FBC004 and could stimulate the production of bacterial exopolysaccharides, resulting in the binding together of soil particles in the vicinity of bacteria (35, 41); and (ii) the absence of plant root and of soil microfauna (e.g., earthworms) which are known to embed bacteria in soil aggregates (10).

In addition to these differences of location with respect to soil porosity, the analysis of presence frequencies in microsamples indicates that introduced bacteria were less dispersed than the indigenous communities for similar abundances (Fig. 1). The estimated cooccurrence probability of introduced FBC004 and indigenous 2,4-D degraders in size-50 microsamples was low when the indigenous 2,4-D degraders were at an abundance level of 10² CFU g⁻¹ (Table 3). This illustrates how much the spatial distribution of bacteria may limit the contact between introduced and indigenous bacteria and thus horizontal gene transfers between them. Nonetheless, it is possible to enhance the probability of cooccurrence of introduced FBC004 and indigenous 2,4-D degraders by manipulating their spatial distribution. Indeed, cooccurrence probability could theoretically be increased by at least 2 orders of magnitude, reaching 10⁴ to 10⁶ cooccurrences g⁻¹, by a percolation of benzoate and 2,4-D aqueous solutions (Table 3). We cannot completely rule out the possibility that the bacterial spatial distributions and thus the cooccurrence probabilities could be different in undisturbed soil. Nevertheless, the fact that a 2,4-D amendment of undisturbed soil provokes a spatial spreading of the indigenous 2,4-D degraders (5) suggests that the mechanisms described here are general.

Comparing the cooccurrence probability estimates with the transconjugant abundances reported in the literature is difficult. First, the simultaneous presence of the two bacterial types in a size-50 microsample, which was evaluated here, is far from guaranteeing a contact between them. Second, once the contact is established, the conjugation will only occur depending upon the capacity of the plasmid for transfer, termed fertility (25). Third, if indigenous transconjugant abundances are measured several days after donor introduction, it is impossible to distinguish the first-generation transconjugants, which received the plasmid through conjugation with the donor, from the bacteria which resulted from the subsequent growth of this first generation. Among the studies which attempted to estimate indigenous transconjugant abundance within 24 h after the donor introduction, which would constitute a fairly good estimation of the first-generation abundance, many failed to detect transconjugant (7, 8, 43, 44). When transconjugants were found, their densities were low (ca. 10 CFU g⁻¹ soil) (44) or the probability of mating occurring on the agar spreading plates could not completely be ruled out (30). Altogether, this suggests that the formation of first-generation transconjugants is a rare event (usually within the range of detection limits), in accordance with the low contacts probability estimated in this work. Indigenous transconjugants are usually unambiguously detected after the first few days following donor introduction. This time frame would allow the donors to spread and form new transconjugants and the first-generation transconjugants to grow and/or transfer the plasmid to other indigenous bacteria.

This microscale study of the spatial distribution of introduced bacteria in soil provided results that can be applied to the various processes which require contact between introduced cells and any component of the soil ecosystem. They indicated that immediately after its introduction, an introduced bacterial population occupies a limited volume of soil, compared to the two indigenous communities studied here. Moreover, increasing the density of the inoculum does not result in a more widespread distribution of the introduced population. Nevertheless, the spatial distribution of both introduced and indigenous bacterial populations can rapidly (within a week) be modified by amending soil with specific substrates. In the case of P. putida (this study) and indigenous 2,4-D degraders (34), a significant spreading of cells within the soil matrix was obtained. This manipulation of bacterial spatial distribution through the modification of environmental conditions offers a promising opportunity to more efficiently use bioengineered bacterial strains in complex environments by increasing their potential contact with their targets. Also, based on estimations of contact, more pertinent risk assessment and modeling of the fate of bacteria or pollutants in soil may be achieved. These results clearly justify the spatial approaches at the microscale, which seems fundamental to the understanding of bacterial processes in the soil environment, whether they are biochemical or genetic processes. It should then be evaluated if this scale of study is more pertinent to modeling and prediction of these processes than the larger scale usually used.