Plasmid Introduction in Metal-Stressed, Subsurface-Derived Microcosms: Plasmid Fate and Community Response

Abstract

The nonconjugal IncQ plasmids pMOL187 and pMOL222, which contain the metal resistance-encoding genes czc and ncc, were introduced by using Escherichia coli as a transitory delivery strain into microcosms containing subsurface-derived parent materials. The microcosms were semicontinuously dosed with an artificial groundwater to set a low-carbon flux and a target metal stress (0, 10, 100, and 1,000 μM CdCl₂), permitting long-term community monitoring. The broad-host-range IncPα plasmid RP4 was also transitorily introduced into a subset of microcosms. No novel community phenotype was detected after plasmid delivery, due to the high background resistances to Cd and Ni. At fixed Cd doses, however, small but consistent increases in Cd^r or Ni^r density were measured due to the introduction of a single pMOL plasmid, and this effect was enhanced by the joint introduction of RP4; the effects were most significant at the highest Cd doses. The pMOL plasmids introduced could, however, be monitored via czc- and ncc-targeted infinite-dilution PCR (ID-PCR) methods, because these genes were absent from the indigenous community: long-term presence of czc (after 14 or 27 weeks) was contingent on the joint introduction of RP4, although RP4 cointroduction was not yet required to ensure retention of ncc after 8 weeks. Plasmids isolated from Ni^r transconjugants further confirmed the presence and retention of a pMOL222-sized plasmid. ID-PCR targeting the RP4-specific trafA gene revealed retention of RP4 for at least 8 weeks. Our findings confirm plasmid transfer and long-term retention in low-carbon-flux, metal-stressed subsurface communities but indicate that the subsurface community examined has limited mobilization potential for the IncQ plasmids employed.

One of the most significant types of soil and subsurface contamination is heavy metal contamination. The bioavailable portion of a heavy metal (i.e., the part that causes the biological response) is typically only a fraction of the total heavy metal concentration in a soil or subsurface environment and may be time variant (1, 25, 38), yet some generalities on the effects of different concentrations of heavy metals can be made (57). At very high levels of heavy metal contamination (e.g., >1% by weight), most bacterial life is inhibited, community diversity is severely reduced, and only a very limited, extremely resistant fraction survives. At lower concentrations (approximately 1 to 10 mM individual metals), microbial activity is temporarily impaired, but acquisition and expression of heavy-metal resistance genes can occur, and the main events in community dynamics may be the spread of metal resistance genes, often plasmid or transposon encoded (26, 34), and outgrowth of the intrinsic heavy-metal-resistant populations (1, 22). At heavy metal levels at or below toxic concentrations (<1 mM), the changes in community dynamics have not been well documented.

This study focused on one component of microbial community dynamics in the subsurface with heavy metal contamination at or below toxic bioavailable concentrations: horizontal gene transfer. It has been posited that a significant contribution to short-term adaptation of microbial communities may be the horizontal, or lateral, exchange of genetic information among their members (50, 56). The higher incidence of plasmids, a dominant form of horizontally transmissible DNA, in bacterial strains isolated from polluted sites than in those from pristine sites has suggested the importance of horizontal gene transfer in polluted environments (6, 35). Hence, we examined the incidence of horizontal gene transfer in subsurface microbial communities that had historically been exposed to subtoxic concentrations of heavy metal. Because of its assumed importance in environmental settings (55), the focus was on conjugal plasmid transfer. Although examination of gene flow in natural microbial communities is experimentally difficult, several studies have demonstrated the occurrence of horizontal gene transfer in a variety of settings such as fresh waters (3, 21) and soils (24, 28, 32). However, essentially no studies on horizontal gene transfer in subsurface microbial communities have been reported. Several reports have suggested that horizontal gene transfer in the environment occurs at lower rates than those observed in laboratory broth mating environments (4, 32, 40), and hence the incidence of gene transfer in the subsurface, with limited carbon fluxes and specific growth rates, may be small (14, 46, 47). Further, it is not clear how the rate and extent of horizontal gene transfer may be affected by the degree of environmental stress, such as the presence of heavy metals.

Our objective, therefore, was to directly examine the incidence of conjugal plasmid transfer and survival in subsurface-derived microcosms operated under conditions designed to mimic subsurface carbon fluxes and subject to different levels of heavy metal stress. The microcosms derived from a subsurface environment with historic heavy metal exposure. The plasmids introduced, pMOL187 and pMOL222, are both pKT240 (IncQ Tra⁻) derivatives containing the czc and ncc operons, which encode high-level resistance to Cd, Zn, and Co and to Ni, Cd, and Co, respectively (27, 44). Both plasmids are nonconjugal, and the transfer of the nonconjugal pMOL222 and pMOL187 plasmids to the indigenous community is contingent on mobilization by a cointroduced mobilizing plasmid or on retromobilization by the indigenous community. Hence, the ability of the community to capture these plasmids (via retromobilization) was compared with its ability to retain the plasmids after cointroduction of an exogenous conjugal plasmid (via mobilization).

MATERIALS AND METHODS

Bacterial strains and plasmids.

Escherichia coli strains DH10B and DH5α were used to deliver plasmids pMOL187 and pMOL222 and plasmid RP4, respectively. Plasmid pMOL187 is a pKT240 derivative (IncQ Tra⁻ Mob⁺ Ap^r Sm^r Kan^r) (2) in which a HindIII-SstI fragment containing the czcICBAD genes (which encode resistance to Cd, Zn, and Co [27]) replaces the original plasmid HindIII-SstI fragment, resulting in loss of the plasmid's Km^r and Sm^r markers. pMOL222 is a pKT240 derivative that has a BamHI fragment containing the nccYXHCBAN and nreAB genes (which encode Ni, Cd, and Co resistance [44]) cloned into the plasmid's unique BamHI site and that retains all the plasmid markers. pMOL plasmids were gifts from M. Mergeay (SCK/CEN, Mol, Belgium). RP4 is the archetypal IncPα plasmid (Tra⁺ Km^r Tc^r Ap^r) (52).

Growth conditions and media.

E. coli strains were grown and enumerated on Luria-Bertani (LB) media with appropriate selective agents (ampicillin [AMP] at 100 μg/ml, tetracycline at 10 μg/ml, and kanamycin at 30 μg/ml) and were incubated at 37°C. Indigenous soil bacteria were grown and enumerated on a 10-fold-diluted R2A medium (1R2A; Difco, Detroit, Mich.). Heavy metal resistance was monitored by enumeration on 0.1R2A agar amended with 10 mM CdCl₂ (R2A-Cd) or 5 mM NiCl₂ (R2A-Ni). All enumerations were conducted by drop plating (29). A solution of 10 mM MgSO₄ was used for all dilutions and resuspensions. An artificial groundwater (AGW), patterned after the existing groundwater, contained 0.5 mM NaHCO₃, 7 mM CaCl₂, 0.15 mM NaH₂PO₄, 690 mg of MgSO₄ · 7H₂O/liter, 5.5 mg of KCl/liter, 1.85 mg of AlCl₃ · 6H₂O/liter, 0.4 mg of NH₄Cl/liter, and SL7 trace mineral stock solution (25% HCl with the following reagents, in milligrams/liter: ZnCl₂, 70; MnCl₂ · 4H₂O, 100; H₃BO₃, 60; CoCl₂ · 6H₂O, 200; CuCl₂ · 2H₂O, 20; NiCl₂ · 6H₂O, 20; NaMoO₄ · 2H₂O, 20) at 1 ml/liter. The AGW was supplemented with 1.0% tap water, and peptone (10 mg/liter) was added as a surrogate carbon source. The AGW was amended with 1, 10, 100, or 1,000 μM CdCl₂ in different treatments.

PCR-based detection of plasmids and donor strains.

The czc operon was detected as a 404-bp czcD gene fragment (GenBank accession no. X98451; primers CzcDUpp [5′-GGACATTTGGCTACTACCG-3′] and CzcDLow [5′-TTCGAGCAGCACATTCAGG-3′]). The ncc operon was detected as a 621-bp fragment spanning the nccA and nccN genes (GenBank accession no. L31363; primers NccUpp [5′-GTGGCGTGTTTGCTCTGGTGT-3′] and NccLow [5′-GGTGTCATCCCATTGCGAGTG-3′]). RP4 was detected by employing IncP rep-specific primers that target an 889-bp fragment of the replication function trafA (17) (IncP-Upp [5′-GTCGAAGCCGTGTGCGAGA-3′] and IncP-Low [5′-GACAGGTTAGCGGTGGCCGAA-3′]). E. coli strains were identified based on a 967-bp fragment of the rrsA3 gene of E. coli (GenBank accession no. MG1655; primers Ecoli-Upp [5′-TCGAACGGTAACAGGAAGAAGC-3′] and Ecoli-Low [5′-GGCACATTCTCATCTCTGAAAA-3′]). A check for nonspecific binding in the Ribosomal Database Project (http://rdp.cme.msu.edu/html/analyses.html) revealed that the E. coli-targeted primers were specific not to E. coli but to Enterobacteriaceae, which was still deemed sufficient. The specificities of other primers to their targets were verified by using the National Center for Biotechnology Information (NCBI) BLAST search engine (http://www.ncbi.nlm.nih.gov/BLAST). PCR conditions for all primer sets were as follows: for czcD, a 5-min denaturation at 94°C, 30 cycles of 1 min at 94°C, 1 min at 56.6°C, and 3 min at 72°C, and a final 7-min elongation at 72°C; for nccAN, a 5-min denaturation at 94°C, 25 cycles of 1 min at 94°C, 1 min at 60°C, and 3 min at 72°C, and a final 7-min elongation at 72°C; for IncP-trafA1, a 10-min denaturation at 94°C, 30 cycles of 1 min at 94°C, 1 min at 57°C, and 3 min at 72°C, and a final 7-min elongation at 72°C; for E. coli rrsA3, a 5-min denaturation at 94°C, 29 cycles of 30 s at 94°C, 30 s at 62°C, and 30 s at 72°C, and a final elongation of 7 min at 72°C.

Infinite-dilution whole-cell PCR (ID-PCR) was used to quantify ncc, czc, trafA, and rrsA3 copy numbers in microcosm eluates (37). This method involved making duplicate sequential dilutions of the template sample (microcosm eluate), performing standard PCRs with standard volumes of the template dilution series (37), and scoring the highest dilution yielding the correct amplicon. PCR products were separated via agarose gel electrophoresis (25 cm, 2% agarose, at 200 V for 2 h) and scored after standardized ethidium bromide (0.5 μg/ml) staining and image acquisition (Gel-Doc 1000; Bio-Rad). The detection limit was calculated as the gene copy number present in the highest dilution that still yielded a detectable amplicon. Detection limits (in gene copy number per PCR) were calculated from a dilution series of the various E. coli host strains carrying the different plasmids in a background of approximately 10⁸ cells/ml, with assumed copy numbers per cell of 10 for the pKT240-derived plasmids pMOL187 and pMOL 222 (15), 5 for RP4 (19), and 7 for the 16S rRNA genes in the E. coli genome (5). The resulting detection limits were 1,200 ± 800, 95 ± 15, 60 ± 4, and 1,300 copies of czc, ncc, trafA, and E. coli rrsA3, respectively.

Colony hybridizations.

Colonies were transferred to a nylon membrane (Immobilon-Ny+; Millipore, Bedford, Mass.) by using standard colony lift and denaturation procedures. Detection was conducted with the colorimetric DIG nucleic acid detection kit according to the manufacturer's instructions by using a digoxigenin-UTP-labeled probe and stringent hybridization conditions (Roche Diagnostics, Indianapolis, Ind.).

Microcosm construction and operation.

Two sets of microcosm studies using identical parent materials were conducted. Sediment samples were retrieved (December 1997), by manual split-spoon sampling techniques, from 8 ft below ground surface (just below the phreatic line), at a site historically contaminated with heavy metals due to the adjacent waste lagoon of a metal plating facility. Samples were kept on ice during transportation to the lab and were stored at −80°C before use. Microcosm series 1 was initiated in October 1999, and series 2 was initiated in June 2000. The total metal concentrations, determined after hydrofluoric acid extraction, were 79 μg/g (0.71 mmol/kg) of Cd²⁺, 96 μg/g (1.64 mmol/kg) of Ni²⁺, and 63 μg/g (0.96 mmol/kg) of Zn²⁺. The NaNO₃-exchangeable metal concentrations, which provide some indication of bioavailable fractions, were determined as 35 μg/g (0.31 mmol/kg) of Cd²⁺, 8 μg/g (0.14 mmol/kg) of Ni²⁺, and 5 μg/g (0.08 mmol/kg) of Zn²⁺ by following the Tessier method (51). The geology of the site consisted of glacial lacustrine deposits of sands and fine sands overlying till and bedrock, with aquifer flow rates varying from 1.48 to 15.35 ft/day. The solid organic carbon content (f_oc, mass of organic carbon per mass of solids) at this depth varied from 0.12 to 0.33%. All analyses were conducted at the Environmental Research Institute of the University of Connecticut.

The microcosms were constructed in 40-ml capped vials (with sterile air vents), operated in aseptic mode, and maintained at 20°C. Five grams of sediment was placed on top of a layer of sterile glass beads covered with a thin layer of glass wool to prevent seepage of the soil particles into the glass beads. Each microcosm had separate sampling and feeding ports. Sediment for control microcosms was sterilized by autoclaving twice in a 24-h period (at 121°C for 20 min). The microcosms were incubated for 3 days with 5 ml of AGW, and the bacterial population was enumerated immediately before inoculation with plasmid delivery strains. At the time of inoculation, 2 ml of seeped liquid was extracted from the glass bead layer and replaced with 2 ml of the appropriate inoculum suspension in AGW. The inocula consisted of E. coli strains carrying pMOL187 (series 1) or pMOL222 (series 2), with or without the addition of E. coli carrying RP4. Microcosm effluent was sampled and replaced at weekly intervals, by drawing 2 ml of percolated water from the glass bead layer and applying 2 ml of AGW on top of the soil through the influent port. Each treatment was performed in duplicate (Table 1).

TABLE 1.

Experimental treatments applied to different microcosms^a

Microcosm	Soil autoclaved?	pMOL187 or pMOL222 introduction?	RP4 introduction?	Cadmium concn (μM)
1-4	N	Y	N	0, 10, 100, 1000
5	N	N	N	0
6	Y	N	N	0
7	Y	Y	N	0
8-11	N	Y	Y	0, 10, 100, 1000
12	Y	Y	Y	0
13-15b	N	N	N	10, 100, 1000

^aY, yes; N, no.

^bOnly for the second microcosm series.

Evaluation of overall community heavy metal resistance.

Overall resistance to cadmium in microcosm series 1 was evaluated by placing 20-μl drops of CdCl₂ solutions (0.1 to10 mM) at distinct locations on a 0.1R2A plate spread with an aliquot of microcosm eluent large enough to produce a complete lawn and scoring growth in and around the drops after 2 weeks of incubation. Overall resistance to nickel in the second microcosm series was evaluated by depositing 20 μl of an NiSO₄ solution on a paper disk (diameter, 13 mm) placed at the center of an R2A plate and scoring growth for proximity to the disk (30). Isolated colonies growing near the filter edge were repatched to media supplemented with Ni (7.5 and 10 mM). To confirm that isolates were not surviving E. coli cells, growth on LB medium at 42°C was evaluated.

Statistical analyses.

To evaluate the effects of various treatments (plasmid and/or metal additions), appropriate 1-way or 2-way analysis of variance (ANOVA) tests were performed. Raw data (each data point is the average of at least triplicate plate counts from a single dilution series) were log transformed to ensure error normality (49). For microcosm series 1 and 2, each experimental treatment resulted, on average, in 22 and 6 data points, respectively. Homogeneity of variances was checked via Hartley's F_max test at 0.05% confidence (49). In the case of a significant treatment effect (at a P value of <0.05), the effect of plasmid additions (RP4 and/or metal resistance plasmids) on means was evaluated by comparing the difference among paired means with the least significant difference, computed as , where df is the degrees of freedom of the error, MS_error is the mean square of the error, and n is the number of observations per group (49).

RESULTS

Fate of introduced E. coli strains.

The population sizes of the E. coli strains introduced (containing RP4 or one of the IncQ plasmids) decreased rapidly in the presence of the indigenous soil community. The strains introduced in the first microcosm series were no longer detectable in the eluates 5 weeks after inoculation when bacteria were enumerated by selective plating (Fig. 1), while they persisted in the autoclaved controls. ID-PCR confirmed that E. coli survived only in autoclaved control microcosms (in series 1) up to 27 weeks after inoculation (between 1.08 × 10⁵ and 7.2 × 10³ E. coli 16S ribosomal DNA templates/ml of microcosm eluate), while positive signals were never found in the noninoculated microcosms. Enumeration of E. coli was not performed in the second series.

FIG. 1.

Fate of introduced E. coli strains in microcosm series 1. Cell numbers are based on enumeration on LB medium plus AMP (100 mg/liter) with incubation at 42°C. The indigenous community was present unless otherwise noted. Symbols: ○, 0 μM Cd; ⋄, 10 μM Cd; □, 100 μM Cd; ▿, 1,000 μM Cd. Filled symbols, only E. coli(pMOL187) was introduced; open symbols, both E. coli(pMOL187) and E. coli(RP4) were introduced; dashed lines, microcosms were autoclaved before strain introduction.

Total-community phenotypic evaluation.

To evaluate whether the mode of microcosm operation (semicontinuous supply and withdrawal of an AGW with a low concentration of a carbon source) allowed a pseudo-steady-state community density, cells in eluates were enumerated on a nonselective medium (0.1R2A) throughout the operation (Fig. 2). The profiles did not differ internally and did not reveal any obvious trends with time, except for an apparent decrease in the total number of cells at the highest Cd dose used. By comparing the numbers of cells retrieved from microcosms 1, 5, and 8, a single-factor-classification ANOVA was first performed to assess whether, in the absence of Cd stress, the addition of plasmid-containing strains affected the number of cells retrieved on nonselective or selective media. Plasmid introduction had no effect on numbers of cells retrieved on R2A (F = 1.04, P = 0.36) or R2A plus Cd (F = 1.66, P = 0.20), while the numbers of cells retrieved on R2A plus AMP plus Cd were significantly affected by plasmid introduction (F = 7.27, P = 0.005), with consistently higher numbers of cells for microcosms subjected to plasmid introduction (microcosms 1 and 8). Subsequently, a two-factor-classification ANOVA was performed to assess whether either the metal dose or plasmid introduction had a significant impact on the total number of cells recovered in microcosm eluates.

FIG. 2.

Cell densities (means ± standard deviations) obtained from direct enumeration of microcosm eluates (series 1) on 0.1R2A. (A) Only E. coli(pMOL187) was added. (B) Both E. coli(pMOL187) and E. coli(RP4) were added. Symbols: ○, 0 μM Cd; ▿, 10 μM Cd; □, 100 μM Cd; ⋄, 1,000 μM Cd; ▵, no strain introduction.

Neither the addition of metal at increasing concentrations nor the introduction of RP4 had a significant effect on the total number of cells throughout the duration of experiment 1 (Fig. 2). To evaluate whether the addition of heavy metals or of plasmids encoding heavy metal resistance resulted in an observable community phenotype, cells in the microcosm eluates were enumerated on a metal-supplemented medium (0.1R2A plus 5 mM Cd) (Fig. 3A and B). The profiles, surprisingly, were very similar for several weeks of operation, although in the latter part of the experiment the different microcosms responded differently. Statistical analysis was performed to assess whether metal dose or plasmid introduction had any effect on the metal-resistant community fraction observed throughout the entire microcosm operation or the latter part of operation.

FIG. 3.

Cell densities (means ± standard deviations) obtained from direct enumeration of microcosm eluates (series 1) on 0.1R2A plus 5 mM Cd (A and B) or 0.1R2A plus 5 mM Cd plus 100 mg of AMP/liter (C and D). Either E. coli(pMOL187) alone (A and C) or both E. coli(pMOL187) and E. coli(RP4) (B and D) were added. Symbols: ○, 0 μM Cd; ▿, 100 μM Cd; □, 100 μM Cd; ⋄, 1,000 μM Cd; ▵, no strain introduction.

The number of Cd-resistant cells was significantly affected by the addition of RP4 (in addition to pMOL187) and the addition of Cd stress, while there was no discernible interaction effect (Table 2). The effects obtained were statistically most significant (smaller P values) when the data of the latter period of operation were evaluated. Because the introduced pMOL187 plasmid encodes both cadmium and AMP resistance, the fraction resistant to both Cd and AMP was also enumerated (Fig. 3C and D). Similar results were obtained. In addition, the introduction of either RP4 alone or RP4 plus pMOL187 enhanced the density of Cd- plus AMP-resistant cells (Table 2). Further inspection revealed that the addition of RP4 consistently increased the numbers of cells as enumerated on media containing either Cd alone or Cd plus AMP (at all Cd concentrations used), and this effect was statistically most significant at the highest Cd concentrations (Table 3).

TABLE 2.

Two-way ANOVA of community phenotypic counts in microcosm series

Microcosms	Resulta for the indicated medium and period
	R2A, entire period			R2A + metal						R2A + metal + AMP, last 12 wk
				Entire period			Last 10 wk
	F	Fcrit	P	F	Fcrit	P	F	Fcrit	P	F	Fcrit	P
Series 1b
RP4 effect	3.53	3.89	0.06	3.69	3.90	0.06	12.21	3.97	8.18 × 10−4	17.34	3.94	6.82 × 10−5
Metal effect	0.66	2.66	0.58	11.41	2.66	7.51 × 10−7	23.60	2.73	9.62 × 10−11	15.84	2.70	1.92 × 10−8
RP4-metal interaction effect	0.63	2.66	0.60	0.27	2.66	0.85	0.85	2.73	0.85	2.15	2.70	0.10
Series 2c
Plasmid effect	33.85	3.15	1.44 × 10−10	15.51	3.15	3.72 × 10−6		ND			ND
Metal effect	19.26	2.76	7.28 × 10−9	6.85	2.76	4.83 × 10−4
Plasmid-metal interaction effect	2.65	2.25	0.01	2.50	2.25	0.03		ND			ND

^aAll results are based on a log transformation of cell counts. Boldfaced values indicate significant effects. ND, not done.

^bRP4 plasmid addition and Cd addition.

^cPlasmid addition and Ni addition; analysis for weeks 6, 7, and 8.

TABLE 3.

Effects of plasmid addition on community resistance phenotypic counts in microcosm series

Cd treatment (μM)	Log-transformed counts of cells with resistance to the following substance(s) after introduction of the indicated plasmid(s):
	Cda			Cd + AMPb			Nic
	pMOL187	pMOL187 + RP4	LSDd	pMOL187	pMOL187 + RP4	LSD	None	pMOL222	pMOL222 + RP4	LSD
0	6.00	6.29	0.41	6.51	6.72	0.31	3.55	5.36e	4.16	1.27
10	6.15	6.32		6.73	6.88		3.24	4.21	4.40
100	6.19	6.83f		6.79	7.45f		3.42	4.29e	5.60e,f
1,000	7.10	7.57f		7.16	7.44f		1.58	2.75e	4.54e,f

^aMicrocosm series 1, last 10 weeks.

^bMicrocosm series 1, last 12 weeks.

^cMicrocosm series 2, weeks 6 to 8.

^dLSD, least significant difference at an α value of 0.05.

^eSignificantly different from microcosms with no plasmid introduction.

^fSignificantly different from microcosms with introduction of a pMOL plasmid but no RP4.

The second microcosm series was conducted for a much shorter period (8 weeks), and enumeration of the metal-resistant fraction (growing on 0.1R2A plus 10 mM Ni) was performed only during weeks 6 through 8 (Table 4). Hence, the effects of plasmid introduction and heavy metal concentrations were evaluated by using measurements for duplicate microcosms at three time points (yielding six observations per treatment)

TABLE 4.

Cell densities obtained from direct enumeration of series 2 microcosm eluates on R2A medium with or without Ni

Plasmid introduced	Cd concn (μM)	Cell densitya (104 CFU/ml) on the following medium at the indicated time:
		0.1R2A			0.1R2A + 5 mM Ni
		Wk 6	Wk 7	Wk 8	Wk 6	Wk 7	Wk 8
None	0	28.6 ± 3.2	19.2 ± 12.3	14.8 ± 14.9	0.150 ± 0.071	0.115 ± 0.134	5.40 ± 3.11
	10	1.74 ± 1.33	5.20 ± 6.79	1.72 ± 2.02	1.22 ± 1.69	1.22 ± 1.69	0.289 ± 0.381
	100	1.90 ± 0.71	1.10 ± 0.14	0.627 ± 0.440	0.609 ± 0.0717	0.609 ± 0.0717	0.0500 ± 0.0141
	1,000	0.0501 ± 0.0706	0.0101 ± 0.0141	0.0708 ± 0.0999	0.0949 ± 0.134	0.0448 ± 0.0632	0.0837 ± 0.118
pMOL222	0	201 ± 76.9	192 ± 164	51.0 ± 43.8	70.2 ± 0.33	23.7 ± 15.1	10.9 ± 10.5
	10	46.7 ± 14.4	353 ± 463	27.1 ± 7.3	2.53 ± 2.73	1.67 ± 2.04	3.20 ± 0.28
	100	86.8 ± 57.1	23.7 ± 33.5	40.6 ± 21.7	33.6 ± 7.6	2.60 ± 3.68	10.5 ± 5.0
	1,000	2.98 ± 0.52	5.70 ± 5.23	2.57 ± 0.61	0.130 ± 0.184	0.960 ± 1.33	0.802 ± 1.11
pMOL222 + RP4	0	431 ± 60.7	322 ± 83.1	127 ± 22.7	1.78 ± 1.69	2.78 ± 3.14	1.71 ± 1.38
	10	99.8 ± 80.3	55.7 ± 57.6	39.7 ± 35.1	1.84 ± 1.42	3.70 ± 3.20	4.62 ± 4.42
	100	162 ± 119	199 ± 232	72.4 ± 91.2	187 ± 135	97.0 ± 126	30.8 ± 41.3
	1,000	16.9 ± 4.5	15.5 ± 10.1	6.91 ± 4.38	14.3 ± 18.7	12.1 ± 11.2	1.20 ± 1.23

^aValues are means±standard deviations.

Both the addition of metal at increasing concentrations and the addition of RP4 had significant effects on the total number of cells (as determined by plating on 0.1R2A) during this short-term experiment (Table 4), chiefly resulting in lower numbers of cells at increasing Cd concentrations, an effect mitigated by the addition of plasmids. Statistical analysis was performed to assess whether the metal dose or plasmid introduction had any effect on the metal-resistant (Ni^r) community observed during the short-term operation (Tables 2 and 3). Again, both plasmid introduction and metal addition affected the Ni^r cell densities retrieved. Because side-by-side experiments were conducted in which pMOL222 was introduced either alone or together with RP4, the effect of either plasmid on the resulting community phenotype could be assessed. Clearly, introduction of pMOL222 by itself increased the Ni^r fraction at elevated Cd concentrations, but joint introduction of RP4 significantly enhanced this resistant population (Table 3).

Because single-point estimates of community metal resistance did not yield the expected phenotypic distinction between different treatments, more-refined estimates were obtained. Both methods permit comparison of a test microcosm's microbial community to a reference microbial community, but neither permits absolute estimates of degree of metal tolerance.

However, whole-cell hybridizations—using the nccAN amplicon as a probe—revealed that Ni^r isolates retrieved from the series-2 microcosms with plasmid introduction contained an ncc gene copy. A total of 71, 196, and 141 isolates were obtained from microcosm sets 13 to 15 (no plasmid introduction), 1 to 4 (pMOL222 introduction), and 8 to 11 (pMOL222 and RP4 introduction), respectively. From these respective microcosm sets, 0, 57, and 20% of the total Ni^r isolates were ncc⁺, while 0, 100, and 86% contained a plasmid consistent with the size of pMOL222. Multiple plasmids (of different sizes) were found in 17, 0, and 7% of the respective microcosm sets. Because none of the isolates were able to grow at 42°C, they were considered non-E. coli.

Direct measurement of plasmid-borne genes in microcosm eluates by ID-PCR.

Larger copy numbers (10- to 1,000-fold) of czc and ncc were consistently found in microcosms that received both the metal resistance and RP4 plasmids (series 1) than in those that received only the metal resistance plasmids (Fig. 4). Neither czc nor ncc gene copies were ever detected in microcosms that were not inoculated with the heavy metal resistance plasmids (data not shown). Irrespective of the metal dose, copy numbers of the metal resistance genes were highest early after donor introduction (weeks 2 and 3 for pMOL187 and pMOL222, respectively), potentially due to survival of donor organisms (Fig. 4). However, even at this point, detection of czc seemed dependent on the presence of RP4. During later enumerations (weeks 14 and 27 for pMOL187), RP4 appeared requisite for detection of czc signals, and czc was detected in only one of the microcosms to which RP4 was not added (one of the duplicate 10 μM Cd treatments). However, at 8 weeks of incubation, the density of ncc in microcosm series 2 was not yet contingent on the copresence of RP4, although a substantially higher ncc copy number was found at the highest Cd doses. The densities of ncc in the control microcosms (microcosms 7 and 12, which were autoclaved and received E. coli additions) were (9.3 ± 4.1) × 10⁹ and (9.9 ± 5.7) × 10⁷ copies per ml of eluate after 2 and 8 weeks, respectively. With trafA as a target gene, IncP plasmid abundance was determined in series 2 (no such analysis was performed for series 1). IncP copies were detected only in microcosms inoculated with the RP4 plasmid [including control microcosm 12, in which it was present at (5.9 ± 4.1) × 10⁹ and (2.0 ± 2.7) × 10⁸ copies per ml of eluate after 2 and 8 weeks, respectively]; IncP attained high densities after 2 weeks (donor survival) but remained in most microcosms at high levels after 8 weeks (end of experiment).

FIG. 4.

ID-PCR detection of czc, ncc, and trafA in microcosm eluates. (A) Microcosms to which E. coli(pMOL187) was added. (B) Series to which E. coli(pMOL222) was added. E. coli(RP4) was added as indicated. Shaded bars, czc; solid bars, ncc; hatched bars, trafA; asterisks, below detection levels.

DISCUSSION

Although many reports on plasmid transfer in microbial communities exist, they often employ conditions of high cell densities and/or high carbon fluxes, ensuring high specific growth rates. It is well known that high specific growth rates are often associated with high plasmid transfer rates, while the use of high cell densities facilitates the detection of rare conjugation events. It was our intent to study plasmid fate and incidence of plasmid transfer under conditions representative of the subsurface, which typically includes low carbon fluxes. Hence, we worked with sediment cores aseptically retrieved from the subsurface. Although we did not attempt to mimic the actual carbon substrate composition, the total applied carbon flux (in a synthetic groundwater patterned after the aquifer composition) was calculated as approximately 50 mg of carbon/m²/day (based on a typical specific surface area of 0.01 m²/g for fine sand, and expressed with respect to the grain surface area), reflective of highly oligotrophic conditions (42). Our attempt at mimicking the (semi)continuous conditions that are typical of soil and subsurface microbial communities contrasts strongly with other studies, where plasmid transfer in microcosms was examined as a batch process and was monitored for very short periods (see, e.g., references 9, 32, 39, and 54).

To reduce the experimental burden of truly continuously applied fluxes in multiple replicate microcosms, the carbon flux was applied semicontinously by weekly addition and withdrawal of a carbon-supplemented synthetic groundwater. Although this operation undoubtedly resulted in perturbation to the sediment community, a key aim was to attain a steady community density. As judged from total cell counts in the microcosm eluates (Fig. 2), this mode of operation permitted rapid attainment of a nearly steady total cell density.

The use of soil samples derived from an aquifer previously exposed to heavy metals obviously complicated phenotypic evaluations of the plasmids introduced, which contained the complete czc or ncc operon and coded for Cd, Zn, and Co resistance or Ni, Co, and Cd resistance, respectively (11, 44). The Cd (10 mM)- and Ni (5 mM)-resistant fraction (based on enumerations on R2A medium) constituted nearly 25% of the total community density (compare Fig. 2 and 3). The historic contamination resulted in total Cd and Ni concentrations of 0.71 and 1.64 mmol/kg (dry weight [DW]) of soil. Often, no significant effects on soil activity and total cell counts are measured at such concentrations, although they may result in increased metal tolerances in the indigenous community (10, 22, 36) and alterations of community structure (18, 23, 43). Extrapolations from one study to another are difficult, however, because typically the total metal concentrations are reported, and not the bioavailable fractions in either the sample under study or the enumeration medium employed. Nevertheless, typical values for 50% inhibition of metabolic activity for communities isolated from pristine arable soil (with a f_oc of 4.4%) were on the order of 10⁻⁵ and 10⁻⁴ M, while increases in community tolerance (to Cd and Ni) required metal pollution levels above 2 and 6 mmol/kg (DW) of soil (10). In a similar study employing coniferous forest soil (f_oc, 64 to 98%), Cd and Ni doses of 10 and 20 mmol/kg (DW) of soil were inferred to result in tolerance beyond that of pristine soil microbial communities (36). Clearly, the inherent Ni and Cd resistances, observed in our studies, are higher, with approximately 25% of the total community resistant to 10 mM Cd and 5 mM Ni, suggesting the higher availability of the metals in the soils studied given the lower f_oc of approximately 0.12 to 0.33%. Others, furthermore, have observed that the bacteria isolated from soil have an inherent high resistance to heavy metals, regardless of whether the soils were contaminated with metals or not and irrespective of the degree of contamination (1, 13, 45). Notwithstanding the ubiquity of ncc- or czc-related sequences in metal-impacted soil environments (12), PCR-based enumerations did not indicate their presence in the sediment samples studied.

Monitoring plasmid fate and transfer to an indigenous community becomes increasingly complicated when obvious phenotypic assessment is difficult. To avoid interference by the donor strains introduced, it is imperative that they not be able to survive in the microcosm. Although several strains have been constructed to perform the role of a transitory plasmid delivery system (e.g., Pseudomonas putida strains bearing the gef-based suicide cassette [31]), their efficiency in our hands was not acceptable, resulting in survival of escape mutants (C. P. Arango and B. F. Smets, unpublished data). Hence, we used E. coli strains derived from the gastrointestinal tract as plasmid delivery organisms. Clearly, E. coli was unable to establish itself in any of the microcosms in the presence of an indigenous microbial community, as evidenced by selective enumeration (Fig. 1) and ID-PCR targeted to Enterobacteriaceae, and as has often been observed (33, 39, 54).

To examine the possible interaction between plasmid transfer and the degree of metal contamination, increasing Cd doses were applied to the duplicate microcosm (0, 10, 100, or 1,000 μM). The metal doses applied during the study were varied from 1 μmol to 1 mmol of Cd/kg (DW) of soil and hence, at maximum dose, did not exceed the historic degree of metal pollution (0.71 to 0.31 mmol of Cd/kg [DW] of soil, based on total or readily extractable concentrations). Nevertheless, the dissolved-phase Cd concentrations in the soil eluates were significantly higher with higher Cd doses, and some accretion of both total and readily extractable solid-phase Cd occurred over the microcosm operation (Table 5). Hence, large effects on total cell numbers were not expected, although a transitory drop in cell counts was observed at the highest Cd dose used (Fig. 2). Even in previously unexposed soil samples, the effects of Cd doses (at concentrations as high as 40 mmol/kg [DW] of soil) on total cell numbers have been observed as short-term, and continued addition has not resulted in further decreases in total-community density (22).

TABLE 5.

Microcosm effluent and soil cadmium content (series 1)

AGW cadmium concn (μM)	Microcosm no.	Concna (μM) in effluent		Final concna (mmol/kg) in soil at wk 24
		Wk 10	Wk 24	Total	NaNO3 extractable
0	1, 5, 7, 8, 12	6.33 ± 2.9	1.5 ± 2.2	0.62 ± 0.19	0.17 ± 0.09
10	2, 9	10.7 ± 0.8	5.6 ± 1.3	0.91 ± 0.31	0.17 ± 0.12
100	3, 10	38.1 ± 3.6	45.5 ± 15.41	1.37 ± 0.14	0.70 ± 0.21
1,000	4, 11	621.8 ± 40.1	824.2 ± 25.9	5.29 ± 0.39	2.05 ± 1.83

^aMean ± standard deviation.

During microcosm series 1, neither RP4 addition nor Cd addition resulted in significant effects on overall total cell counts in microcosm eluates (Table 2), suggesting minimal disturbance to total-community density. For series 2, however, both plasmid introduction (RP4 or pMOL222) and Cd addition affected total cell counts, with Cd addition causing a decrease and plasmid addition causing a increase in total cell numbers (Table 2). Further, paired comparison of microcosms at a fixed Cd dose indicated that the addition of both plasmids consistently resulted in a higher number of cells than that observed in microcosms without plasmid introduction. It should be recognized that results from microcosm series 2 were comparisons of weeks 6 through 8, while 27 weeks of data were compared for microcosm series 1; hence, the effect was potentially short-term, although no such effect was clearly observed in microcosms series 1. Further, the concentrations used (maximum, 1 mmol/kg of soil) are much lower than those typically considered to have an effect on total-community density (22, 36).

In agreement with the observations on total-community density, the Cd dose did not affect the density of the Cd-resistant population in series 1, although the effect on the Ni^r population in microcosm series 2 was evident: the absolute Ni^r population density decreased with increasing Cd doses, while the fraction of the population that was Ni^r gradually increased with increasing metal doses, irrespective of plasmid addition (14, 17, 34, and 88% at 0, 10, 100, and 1,000 μM Cd, respectively, at 8 weeks). The phenotypic response to Cd doses in terms of Ni resistance is not unprecedented; multiple heavy metal resistance is often noted (10), and multiple resistance has a genetic basis (11, 44).

Typically, in situ plasmid transfer after introduction of an exogenous plasmid-containing strain into soil or sediment microcosms is demonstrated by the identification of a new plasmid-caused phenotype. Such an approach relies on the cointroduction of a well-marked recipient strain (32, 54) or on the appearance of the phenotype in the indigenous community (20, 41). In the latter case, it is again imperative that the original donor not interfere with the enumeration of the new phenotype (i.e., that it either does not survive, can be selectively removed, or does not express the plasmid-encoded phenotype [33]). Although the density of the E. coli donor introduced declined rapidly in all but the autoclaved microcosms, the inherent resistance of the community to Cd and Ni prevented us from scoring the incidence of a new phenotype in the indigenous community as a manifestation of a plasmid effect, even when we attempted to enumerate the indigenous community on media with varying metal concentrations (gradient plates or drop enumerations).

When the effects of plasmid addition and Cd stress on numbers of metal-resistant cells were evaluated, the changes in total cell number that were observed in microcosm series 2 due to Cd additions and those due to plasmid addition had to be separated. Hence, cell numbers in pairs of microcosms that received different plasmids but the same Cd treatment were compared. In microcosm series 1, only the additional effect of RP4 (beyond that of pMOL187) was evaluated: At all doses, the introduction of RP4 resulted in an increased density of the Cd^r or Cd^r-plus-Amp^r population, with statistically significant differences only at the highest Cd doses tested (Table 3). In microcosm series 2, both the addition of pMOL222 and the additional effect of RP4 were tested: at all Cd doses tested, the addition of pMOL222 resulted in an increased Ni^r population, which was consistently further increased by joint introduction of RP4 (Table 3). Again, these effects were statistically most significant at the highest Cd concentrations tested.

Others have suggested the need for pollutant stress to actuate transfer of an exogenously introduced plasmid in a soil microbial community. Often, however, such transconjugant occurrences observed after the application of stress are compounded by the effect of the stress on transconjugant growth, certainly when the stress applied consists of a growth substrate for the plasmid-encoded phenotype (9, 32, 33). Top et al. (54) observed that, in model microcosms (containing autoclaved soil materials) with various introduced donors (E. coli strains) and recipients (Alcaligenes eutrophus strains) at concentrations as high as 10⁸ CFU/g of soil, inclusion of metal (Zn)-contaminated soil resulted in an increased occurrence of transconjugants (in their case, retrotransconjugants that contained a czc-containing plasmid, pMOL155, very similar to those employed in our study), although under the conditions tested, there was no obvious need for the plasmid-encoded genes, since both parental strains were able to survive. They were, however, unable to confirm their findings with nonsterile soil samples, probably due to the low incidence of retrotransfer (54).

Our results suggested that at the highest Cd stresses applied, the introduction of plasmids had an observable effect on the fraction of the community that expressed a phenotype encoded by those plasmids; whether that increase in the phenotypic trend was the direct effect of expression of the genotype introduced could not be verified. However, this increased phenotype was consistently enhanced by the joint introduction of RP4, although RP4 does not code for the phenotype examined. Although RP4 addition was not studied in experimental isolation (addition of RP4 without addition of pMOL plasmids), it is likely that the role of RP4 is indirect, perhaps involving the mobilization or retention of plasmids related to the phenotype examined, since IncP plasmids have a very wide mobilization range, and plasmids in addition to those introduced may have been mobilized (52). Microcosm experiments in which RP4 is introduced as the sole plasmid could provide additional proof of this stimulatory effect.

Notwithstanding our inability to delineate a clear transconjugant population from the microcosms with plasmid introduction, we obtained molecular evidence of the survival and retention of the plasmids introduced, by using plasmid-targeted ID-PCR. Though tedious, this method provides more-accurate estimates of the transconjugant population because it is not culture based, and in that respect it is similar to in situ detection of transconjugants (7).

ID-PCR revealed that both czc (series 1) and ncc (series 2) genes were present in the microcosms at various sampling points. However, long-term presence of czc copies in series 1 (after 14 or 27 weeks) was contingent on the joint introduction of RP4, although RP4 cointroduction was not yet required to ensure retention of ncc in series 2 (after 8 weeks). Furthermore, the RP4-targeted ID-PCR revealed that RP4 was maintained in series 2. Although the enumerations after 2 weeks could reflect survival of the E. coli plasmid donors introduced, ID-PCR with samples from microcosm 12 revealed trafA copy numbers that were 2 orders of magnitude higher than those found in any of the other microcosms, suggesting rapid die-off of E. coli. Plasmid gene signals detected from 8 weeks onward were most likely due to indigenous transconjugant bacteria. Although the positive detection of ncc or czc does not constitute proof that the plasmid is retained, neither the pKT240 plasmid nor the inserted ncc or czc fragments contain known transposable elements, making transposition to indigenous plasmids or chromosomes unlikely and continued presence of the introduced plasmids probable. Further, the elaborate approach employed to retrieve isolates with high Ni^r (series 2) led to the conclusion, inferred from ncc hybridizations and the presence of a pMOL222-sized plasmid, that several strains contained the introduced pMOL222 plasmid.

Our results strongly suggest that the plasmids introduced can be transferred to the indigenous community and retained for prolonged periods, albeit at low densities, even though no clear phenotypic advantage can be ascribed to the plasmid. Although plasmid transfer and mobilization under extreme oligotrophic conditions have been demonstrated (e.g., in rivers and seawater [7]), these experiments are often conducted with very high localized densities of donors and recipients and with broad-host-range conjugal plasmids (16, 21), while transconjugants that are the result of mobilization of nonconjugal plasmids in the indigenous community are often only transiently observed (21, 54). The strong dependency of the copy number on the joint introduction of RP4 strongly suggests that the poor establishment of the pMOL plasmids, when introduced by themselves, was in part due to the indigenous community's limited potential to mobilize these plasmids (53). Although mobilization of IncQ plasmids is not restricted to IncP plasmids (53), the indigenous community did not contain detectable IncP-related plasmids, as evidenced by PCR amplification with the trafA1 primers, which are IncP replicon specific (17). Further, in series 1, we were unable to detect a mobilization potential for IncQ plasmids of the microcosm sample communities by using a standardized assay (53), except for the autoclaved microcosms to which both RP4 and pMOL222 were added (data not shown). These observations, then, may suggest that neither IncP or IncQ plasmids are endemic to the indigenous microflora of these sediments, in agreement with their absence from many other environments (8, 48).

In summary, our work has revealed that exogenously introduced conjugal and nonconjugal plasmids can be retained in subsurface sediment-derived microbial communities subject to degrees of carbon flux typical for subsurfaces and under continued metal stress. These plasmids can be retained in the community even though they do not appear to provide a phenotypic addition to the community, but their survival is contingent on cointroduction of a mobilizing RP4 plasmid. Further, at the highest metal exposures, there was a significant effect of RP4 presence on the measured size of the community fraction expressing the plasmid-encoded phenotype, suggesting that at the highest metal concentrations, plasmid transfer or retention may be enhanced.