Luminescent Gram-negative strains

Escherichia coli-based strains

Escherichia coli is an enteric Gram-negative bacterium with the sequenced genome [25], well developed methods for genetic modification [26] and various laboratory strains available. Although the relevance of this bacterium for environmental testing is questionable, it has been widely used in the construction of various bacterial sensors [9, 10, 12, 14, 17], mainly due to its compatibility with the chemical-response elements used.

In total three new E. coli MC1061 recombinant luminescent sensor strains expressing luxCDABE genes in response to subtoxic concentrations of bioavailable heavy metals ("lights-on" sensors) were constructed: E. coli MC1061(pmerR_BSBPmerlux), E. coli MC1061(pSLcueR/pDNPcopAlux) and E. coli MC1061(pSLzntR/pDNPzntAlux) (Table 1). All the strains harbour plasmid(s) containing metal-response elements from resistance systems for Hg (from Serratia marcescens plasmid [27]; mer), Cu (cue/cop) and Zn (znt) (both from E. coli chromosome [28, 29]) fused with the luxCDABE cassette [see Additional file 1]. MC1061(pSLcueR/pDNPcopAlux) and MC1061(pSLzntR/pDNPzntAlux) contain two plasmids necessary for sensor function, of which one (pSLzntR or pSLcueR, respectively) is present in bacterial cells in high copy number and contains genes for the regulatory proteins (cueRor zntR). The other plasmid (pDNPzntAlux or pDNPcopAlux, respectively) present in medium copy number contains the lux-genes fused to a promoter (P copAor P zntA) controlled by the regulatory proteins. Location of the genes for regulatory protein and the corresponding promoter in different plasmids yields high regulatory protein to promoter ratio necessary to reduce the unfavourably high level of leakage of the P copAand P zntAand thus, high background luminescence of those strains (Table 1). Characteristics (number of cells in the test, background luminescence, maximum induction) of the constructed sensors are presented in Table 1. Induction profiles of those sensor strains with seven tested heavy metals are presented in Figure 1 and resepective limits of determination in Table 1.

	Sensor strains			Control strains
	MC1061 (pmerRBSBPmerlux)	MC1061 (pSLzntR/ pDNPzntAlux)	MC1061 (pSLcueR/ pDNPcopAlux)	MC1061 (pSLlux)a	MC1061 (pDNlux)b
Background luminescence c	230 ± 90	2940 ± 930	520 ± 70	575000 ± 35200	6500 ± 2270
Maximum induction d	400	20	100	no induction	no induction
No of cells in the test (100 μl)	4 × 107	107	8 × 106	8 × 106	107
Compound	Limit of determinatione/Toxicityf, M (mg of compound l-1)			Toxicity (2-h EC50g), M (mg of compound l-1)
HgCl 2	10-10 (0.00003)/5 × 10-6 (1.4)	8 × 10-8 (0.02)/10-5 (2.7)	no induction	10-6 (0.4)	10-6 (0.4)
CH 3 HgCl	8 × 10-12 (0.000002)/5 × 10-8 (0.01)	not tested	not tested	3 × 10-6 (0.7)	not tested
CdCl 2	3 × 10-7 (0.06)/5 × 10-4 (92)	10-8 (0.003)/5 × 10-6 (0.9)	no induction	4 × 10-4 (70)	2 × 10-4 (31)
ZnSO 4 h	no induction	5 × 10-6 (0.8)/5 × 10-2 (8000)	no induction	10-2 (1600)	4 × 10-3 (640)
Pb(NO 3 ) 2	no induction	7 × 10-7 (0.2)/10-3 (330)	no induction	2.4 × 10-4 (79)	6 × 10-4 (190)
CuSO 4 i	not tested	not tested	10-7 (0.02)/10-3 (240)	not tested	5.4 × 10-4 (86)
AgNO 3	not tested	not tested	2 × 10-6 (0.3)/10-5 (2.6)	not tested	2 × 10-6 (0.3)

^a control for MC1061(pmerR_BSBPmerlux)

^b control for MC1061(pSLcueR/pDNPcopAlux) and MC1061(pSLzntR/pDNPzntAlux)

^c luminescence (relative light units, RLU) of 100 μl of bacterial suspension before the test as measured by plate luminometer Fluoroskan (ThermoLabsystems). Average of at least three assays

^d as fold over the (non-induced) background

^e see equation 4 in Methods

^f the lowest tested concentration causing a measurable decrease in the bioluminescence of the sensor strains (read from concentration-induction curves in Figure 1)

^g concentration causing 50% decrease in bioluminescence production (calculated from Figure 2)

^h tested as ZnSO₄ × 7H₂O

ⁱ tested as CuSO₄ × 5H₂O

In parallel with sensors, two E. coli toxicity control ("lights-off") strains: MC1061(pSLlux) and MC1061(pDNlux) were constructed (Table 1). These constitutively luminescent strains lacking metal-response elements contain plasmids in which the luxCDABE genes were either under the lac promoter (pSLlux) or T7 promoter (pDNlux). Indeed, within the limits of error, the control strains were not induced by the heavy metals (Figure 2) and inhibition of bioluminescence (NL) was observed with increasing metal concentrations. Toxicity (2-h EC₅₀ values) of the seven tested heavy metals to E. coli MC1061(pSLlux) and MC1061(pDNlux) are presented in Table 1 and respective concentration-effect curves (used for EC₅₀ calculation) in Figure 2.

Luminescent Gram-positive strains

Bacillus subtilis and Staphylococcus aureus were used to construct Gram-positive heavy metal sensor strains. Bacillus subtilis is a spore-forming bacterium common in soils and thus a relevant environmental test organism. Recombinant B. subtilis has earlier been used as a model Gram-positive bacterium for toxicity testing [35] as well as a host for a recombinant bacterial cadmium sensor [36]. Staphylococcus aureus, although not an environmentally relevant bacterium, it may be a medically important model organism as it is an opportunistic human pathogen often bearing multiple antibiotic resistance.

Two constitutively luminescent Gram-positive control ("lights-off") strains S. aureus RN4220(p602/22lux) and B. subtilis BR151(pBL1/p602/22lux) in which the optimized luxCDABE [37] is under the lac promoter, were constructed (Table 4). However, analogously to earlier studies [38] the stability of the plasmids with genes encoding for bioluminescence was relatively poor in Gram-positive bacterial strains as frequent loss of plasmids especially in strains with higher background luminescence, was observed. Even, no transformants were obtained when plasmid p602/22lux was transformed into B. subtilis BR151. This problem was solved and the relative stability of the bioluminescence encoding genes was achieved when the plasmid was transformed to B. subtilis strain containing a helper plasmid pBL1 (expresses LacIrepressor) [38]. Toxicity (2-h EC₅₀ values) of Hg²⁺, Cd²⁺, Pb²⁺ and Zn²⁺ to S. aureus RN4220(p602/22lux) and B. subtilis BR151(pBL1/p602/22lux) are presented in Table 4 and respective concentration-effect curves in Figure 2.

Properties of "lights-off" control strains

The constitutively luminescent "lights-off" control strains had always higher background bioluminescence than the respective control strains (Tables 1, 2, 3, 4). The highest background luminescence (600 000 RLU) was measured for E. coli MC1061(pSLlux) probably due to the location of the luxCDABE on a high-copy-number plasmid. As the rationale of including the control strains in the assay is to take into account the possible non-specific effects (e.g., toxicity) of the sample on bacterial bioluminescence, the high initial bioluminescence allowing more accurate measurement of the inhibition of bioluminescence, is certainly a bonus. By decreasing the plasmid copy number, the background bioluminescence of the "lights-off" strains clearly decreased. The bioluminescence of E. coli MC1061(pDNlux) containing a medium-copy-number plasmid was about 100-fold lower than that of E. coli MC1061(pSLlux) (Table 1). The same trend was observed for P. fluorescens as the background luminescence of OS8(pDNlux) (luxCDABE in about 10 copies) was about 100-fold higher than that of the OS::Knlux (a single copy of luxCDABE) (Tables 2 and 3). However, the background luminescence value did not affect the sensitivity of the "lights-off" strains to heavy metals: the EC₅₀ values of differently luminescent E. coli and P. fluorescens control strains were similar (Tables 1, 2 and 3). Almost no effect of bioluminescence on bacterial sensitivity to heavy metals has also been demonstrated earlier by showing that a lux-marked Burkholderia sp. had similar sensitivity to Zn and Cu as the wild (not luminescent) strain in dehydrogenase assay [39]. In addition, we showed that the toxicity of Hg, methylmercury, Pb and Ag to the both Gram-negative bacteria (E. coli and P. fluorescens) was similar. However, E. coli proved more sensitive than P. fluorescens towards Zn (about 3-fold), Cu (about 15-fold) and Cd (about 100-fold) (Tables 1, 2 and 3). Indeed, relatively low sensitivity of a Pseudomonas putida towards Cd (about 3 orders of magnitude lower than the sensitivity of E. coli) in a growth ihibition assay has been shown earlier [40].

Comparison of Gram-negative and -positive control strains showed that the luminescence of the latter was always lower (Tables 1, 2, 3, 4). This could be either due to the poor expression of bioluminescence-encoding genes in these strains or partial loss of lux-containing plasmid in the bacterial cultures. In addition, Gram-positive strains were remarkably more sensitive to the tested heavy metals than Gram-negative (EC₅₀ values in Tables 1, 2, 3, 4). Interestingly, other authors [41] have shown that the toxiciy of Hg was similar to E. coli and B. subtilis whereas we showed that the latter was at least 20-fold more senistive than the former.

Properties of heavy metal sensor strains: background luminescence

Differently from "lights-off" constructs, in metal sensor strains the lower background luminescence is an advantage as it ensures a more pronounced induction at lower metal concentrations (in the case of a strain with high background luminescence this high value must be exceeded at least twice to be equivalent to LOD). Theoretically, the sensor strains should be bioluminescent (induced) only in the presence of intracellular heavy metals, which bind on the regulatory protein. However, background luminescence values in the sensor strains were also registered in the absence of added heavy metals (Tables 1, 2, 3, 4). The reason for this could be either the leakage of the promoter controlling the expression of lux-genes and/or as proposed by [42] metal impurities of the test medium. Indeed, the optimized medium used in our study (see Methods), contained trace amounts of Cd (0.0001 mg l^-1), Zn (0.035 mg l^-1), Cu (0.0025 mg l^-1) and Pb (0.003 mg l^-1). However, these concentrations were more than 5-460-fold lower than the LODs of the sensors, depending on the strain (Tables 1, 2, 3, 4). Also, the background luminescence of the sensor strains (varying among the strains; Tables 1, 2, 3, 4) did not correlate with their sensitivities to the metals present as impurities in the test medium. Thus, the most probable reason for the background luminescence in the sensor strains is the leakage of the promoter controlling the expression of luxCDABE genes. Indeed, the E. coli MC1061(pmerR_BSBPmerlux) had remarkably lower background than the other E. coli strains and it is well known that the regulatory protein in this sensor, MerR, is a strong repressor in the absence of mercury [43, 44]. At the same time, the regulatory protein ZntR in MC1061(pSLzntR/pDNPzntAlux) strain with high background luminescence, has not been clearly shown to act as a repressor in the absence of zinc [29]. In addition to the nature of the metal-response elements, also the number of copies in which the lux-cassette and linked metal-response elements were present in the cells affected the background luminescence (P. fluorescens plasmid-bearing versus chromosomal strains, Tables 2 and 3). As a rule, the higher the background luminescence the lower the maximum induction of the sensors (Tables 1, 2, 3, 4). This could be due to the depletion of energy in the bacterial cells at high bioluminescence values as the emission of bioluminescence requires high amounts of cellular energy [45]. However, the background luminescence of the metal sensor strains did not affect their metal sensitivity (LOD) (Tables 1, 2, 3, 4).

Properties of heavy metal sensor strains: sensitivity and specificity

The sensor strains were characterized for their sensitivity and specificity using selected environmentally relevant heavy metals. The induction of luminescence was measured after 2 hours incubation that was selected according to our optimization studies (T.Rõlova, personal communication) and similar studies by other authors [17]. Interestingly, only one of the sensors, P. fluorescens OS8::KncueRPcopAlux was single metal (Cu) specific (Figure 1J). The E. coli MC1061(pSLcueR/pDNPcopAlux) expressing similar metal-response elements was in addition to Cu induced also by Ag, but in relatively high (environmentally irrelevant) concentrations (Figure 1I). The sensors expressing Hg-response elements (E. coli MC1061(pmerR_BSBPmerlux), P. fluorescens OS8(pDNmerR_BSBPmerlux) and OS8::KnmerR_BSBPmerlux)) showed also relatively high specificity, by being induced only with inorganic and organic mercury compounds and Cd. However, induction with the latter occurred at concentrations 3-4 orders of magnitude higher than in the case of mercury and yielded remarkably lower inductions (Figure 1A–C). These data are in accordance with [46–48]. It is interesing to note that E. coli MC1061(pmerR_BSBPmerlux) and P. fluorescens OS8(pDNmerR_BSBPmerlux) were detecting methylmercury at about one order of magnitude lower concentrations than HgCl₂ (Tables 1 and 2). Similar results for the same metal-response elements have been shown in [49] and is most probably caused by higher bioavailability of the organic mercury salt. On the other hand, both inorganic mercury and methylmercury were of similar toxicity to E. coli MC1061(pSLlux) or the latter was even less toxic to P. fluorescens control strains (Figure 2A), indicating that the elevated sensitivity towards methylated mercury occurs only at sub-toxic concentrations.

Sensor bacteria carrying the metal-response elements from Zn-resistance system of E. coli (E. coli MC1061(pSLzntR/pDNPzntAlux) and P. fluorescens OS8::KnzntRPzntAlux), Cd-resistance system from Pseudomonas putida (P. fluorescens OS8(pDNcadRPcadAlux) and OS8::KncadRPcadAlux), Pb-resistance system from Ralstonia metallidurans (P. fluorescens OS8(pDNpbrRPpbrAlux), OS8::KnpbrRPpbrAlux) or Cd-resistance system from S. aureus (S. aureus RN4220(pcadCPcadAlux) and B. subtilis BR151(pcadCPcadAlux)) responded to four of the selected seven heavy metal salts: HgCl₂, CdCl₂, Pb(NO₃)₂ and ZnSO₄ (Figure 1, Tables 1, 2, 3, 4) and the order of induction in all the cases was Hg²⁺ ≈ Cd²⁺ < Pb²⁺ < Zn²⁺. These unspecific responses were somewhat expected as earlier studies have shown similar results e.g. for Zn-resistance determinant from E. coli chromosome [14, 50] and to our knowledge, no specific Cd-response system has been found. Unspecificity of several metal-response systems could be due to the evolution of these metal resistance systems to be induced by several heavy metals in order to detoxify them, or simply because the metal-binding proteins in these systems are unable to differentiate between different bi-valent metal ions [51].

Comparison of sensitivities of the current metal sensor strains with previously constructed ones (see below) showed that some of the current sensors were detecting very low levels of metals (lowest LOD values reported so far). However, variability of test results may occur between different papers due to different experimental conditions, which may affect the speciation and thus, bioavailability of metals. Nevertheless, our comparison showed that the strain E. coli MC1061(pSLcueR/pDNPcopAlux) detected 20-fold lower concentrations of Cu than similar previously published bacterial strains [11] and two strains – E. coli MC1061(pmerR_BSBPmerlux) and B. subtilis BR151(pcadCPcadAlux) – were similar in sensitivity with the most sensitive previously reported strains for Hg and Cd, respectively [10, 11, 16, 52].

Comparison of sensor constructs in different host bacteria

Performance of different types of recombinant bacteria carrying similar metal-response elements was compared. Comparison of the two Gram-negative host bacteria E. coli and P. fluorescens showed that the E. coli-based metal sensors were somewhat more sensitive. For example, E. coli MC1061(pSLcueR/pDNPcopAlux) was about 300-fold more sensitive to Cu than P. fluorescens OS8::KncueRPcopAlux and E. coli MC1061(pmerR_BSBPmerlux) was induced by at least 10-fold lower concentrations of Hg²⁺, methylmercury and Cd²⁺ than P. fluorescens OS8(pDNmerR_BSBPmerlux). This and the lower toxicity of some heavy metals to P. fluorescens compared with E. coli could be explained by different cellular metal homeostasis (e.g., the presence and nature of metal transporters) of the host bacteria. Indeed, the effect of metal transporters on the sensitivity of bacterial heavy metal sensors has been shown [53]. On the other hand, there were no remarkable differences in LOD values of the two Gram-positive sensor strains S. aureus RN4220(pcadCPcadAlux) and B. subtilis BR151(pcadCPcadAlux) carrying similar metal-response elements (Table 4). Thus, differently from "lights-off" strains, the sensitivity of the "lights-on" sensors depends on the nature of the metal sensing element rather than on the bacterial host. Indeed, the strains with metal-response elements from Hg resistance system were the most sensitive to Hg and the least sensitive to Cd regardless of the host strain.

Different location of the metal-response elements in P. fluorescens (on a plasmid present in the cells in about 10 copies or as a single chosomosmal insertion) was chosen to clarify whether lowering the copy number of the metal-response elements and lux genes will yield higher sensitivity, as previously shown for stress-induced bacterial strains [54]. However, our data on P. fluorescens with chromosomal insertions showed lower background luminescence but not higher sensitivity than the respective plasmid-containing strains (Figure 1B, C).

Possible applications

As shown above, all the constructed metal sensor strains showed concentration-dependent linear increase in bioluminescence at sub-toxic concentrations of heavy metals. Moreover, the increased bioluminescence of the sensor implies that the toxic metal has crossed the biological membrane and thus, has a potential to accumulate in the food chain. In addition, the limits of determination of most of the constructed sensors under our test conditions were low enough to detect even the EC permitted limit values for Hg, Cd, Cu and Pb in drinking water (1, 5, 5000 and 10 μg l^-1, respectively; Directive 98/83 EC).

Currently the most challenging field of biosensor applications is the evaluation of bioavailable amounts of heavy metals in complex environmental matrices [55, 56] for risk assessment purposes. However, the use of microbial sensors for environmental analysis is currently restricted mostly due to the lack of specificity. This drawback may be overcome by the combined use of several strains with different specificities and sensitivities. In addition, Cu and Zn-sensors have already proven very useful in other fields, e.g. in quantification of solubilized bioavailable metals from metal-containing nanoparticles [57–59].

Construction of plasmids and strains

Bacterial strains and plasmids used in this study are listed and their detailed characterization is provided in Additional file [see Additional file 1]. All the constructed sensors are based on components of heavy metal resistance systems (metal-response elements) and luxCDABE genes from Photorhabdus luminescens [60] (reporter).

Gram-negative strains

Plasmid-containing strains

Altogether three new broad-host range sensor plasmids were constructed and electroporated into P. fluorescens OS8 [65] by standard methods [26]. The plasmid constructs were verified by multiple restriction analysis.

Plasmid pDNmerR_BSBPmerlux was constructed by digesting the pmerR_BSBPmerlux with PvuII and ligating with BamHI-digested and Klenow-treated pDN18N [64].

Plasmid pDNpbrRPpbrAlux was constructed by ligating the BamHI-XhoI-digested pSLlux [32] and BamHI-XhoI-digested 650 bp PCR product containing pbrR and promoter of pbrA (primers: 5' ATATCTCGAGGGGAGCCTTACCGGCAGAACAGCGA 3'and 5' TTAAGGATCCCCTCATGGCAACCCCTTGTGTGTATTCA 3'; Ralstonia metallidurans CH34 cell extract prepared as in [50] was used as a template). The resulting plasmid pSLpbrRPpbrAlux was digested with PvuII-NaeI and ligated with BamHI-digested and Klenow-treated pDN18N [64] to obtain pDNpbrRPpbrAlux.

Plasmid pDNcadRPcadAlux was constructed by ligating the BamHI-BspEI-digested pSLlux [32] and BamHI-AgeI-digested 550 bp PCR fragment containing cadR and promoter of cadA (primers 5' ATATACCGGTAAGCTTTTGGCTTAATGCCCGTG 3' and 5' TTAAGGATCCCATGGGGTCATCCTTAATTTGAGCC 3' were used; cell extract of Pseudomonas putida PaW85 prepared as in [50] was used as a template). The resulting plasmid pSLcadRPcadAlux was digested with XhoI-SalI, Klenow-treated and ligated with BamHI-digested and Klenow-treated pDN18N [64] to obtain pDNcadRPcadAlux.

The control strain OS8(pDNlux) was from [32].

Strains with chromosomal insertions

Chromosomal insertions were done by Tn5 mini-transposonmutagenesis method [66]. Transposon-carrying suicide plasmids were constructed (see below) and transferred into the recipient strains by conjugation.

Construction of suicide plasmids

All suicide plasmids were constructed in the E. coli strain SY372 λpir [67] and were based on pTCR241 [68]. The pTCR241 was modified as follows: 1500 and 2500 bp NotI fragments between terminal ends of the Tn5 were removed and replaced by kanamycin resistance gene from pET26b(+) (Novagen) and the resulting plasmid was desginated pTCRKn.

For the construction of pTCRKnpbrRPpbrAlux, pTCRKnmerR_BSBPmerlux and pTCRKnzntRPzntAlux, 6900, 8200 and 7000 bp PvuII fragments of pSLpbrRPpbrAlux, pmerR_BSBPmerlux and pSLzntRPzntAlux and for the construction of pTCRKncadRPcadAlux and pTCRKncueRPcopAlux, 6750 and 6500 bp SalI-XhoI-digested and Klenow-treated fragments of pSLcadRPcadAlux and pSLcueRPcopAlux were inserted into StuI-site of pTCRKn situated between the terminal ends of the Tn5 before the kanamycin resistance encoding gene. All the fragments inserted into pTCRKn contained respective metal-response elements and luxCDABE genes. For the construction of pTCRKnlux, an intermediate plasmid was constructed by inserting SalI-NruI fragment of pSLlux (luxCDABE genes under the lac promoter) into SalI-StuI-digested pSL1190. From the resulting plasmid a PvuII fragment (lac promoter and luxCDABE) was inserted into StuI-digested pTCRKn. The plasmid constructs were verified by multiple restriction analysis

Chromosomal insertions to P. fluorescensOS8

For chromosomal insertions, E. coli S17-1 λ pir [69] was transformed with the appropriate suicide plasmid by CaCl₂ method [26] to obtain the donor strains. The donor strain was grown overnight in LB medium (per liter: 10 g of tryptone, 10 g of NaCl, 5 g of yeast extract [26]) with 100 μg ml^-1 of ampicillin and mixed with the overnight culture (in LB) of the recipient P. fluorescens OS8 at a ratio 1:2 (20 μl: 40 μl). 5 ml of 10 mM MgSO₄ was added to the mixture and the cells were collected on a 0.45 μm membrane filter. The filter was incubated on LB agar plate for 8–18 h to allow conjugation after which the cells on the filter were resuspended in 5 ml of 10 mM MgSO₄. 100–500 μl of the resulting mixture was plated onto selective M9 minimal medium (per liter: 6 g of Na₂HPO₄, 3 g of KH₂PO₄, 0.5 g of NaCl, 1 g of NH₄Cl) agar (1.5%) plates supplemented with 4 g l^-1 of glucose and microelements (final concentration per liter: 26.9 mg of MgO, 5 mg of CaCO₃, 0.13 ml of conc. HCl, 11.3 mg of FeSO₄ × 7H₂O, 3.6 mg of ZnSO₄ × 7H₂O, 2.8 mg of MnSO₄ × 4H₂O, 0.65 mg of CuSO₄ × 5H₂O, 0.7 mg of CoSO₄ × 7H₂O, 0.15 mg of H₃BO₃; the mixture was prepared as 400-fold concentrate). Selectivity of the plates was obtained by supplementing the medium with 100 μg ml^-1 of kanamycin. Strains with chromosomal insertions of metal-response elements and luxCDABE genes from pTCRKnpbrRPpbrAlux, pTCRKnmerR_BSBPmerlux, pTCRKnzntRPzntAlux, pTCRKncadRPcadAlux and pTCRKncueRPcopAlux were designated as OS8::KnpbrRPpbrAlux, OS8::KnmerR_BSBPmerlux, OS8::KnzntRPzntAlux, OS8::KncadRPcadAlux and OS8::KncueRPcopAlux, respectively. The strain with chromosomal insertion of luxCDABE from pTCRKnlux was designated as OS8::Knlux.

Gram-positive strains

Plasmids pcadCPcadAlux and p602/22lux were constructed by inserting the SmaI-SpeI-digested 5700 bp luxCDABE from pSB2025 (translational signals before the genes have been optimized for Gram-positive bacteria) [37, 70] into NheI-BamHI-digested (BamHI-digested end was filled with Klenow fragment) pTOO24 [36] or p602/22 [71], respectively. The plasmid constructs were verified by multiple restriction analysis

For the construction of Gram-positive metal sensor and control strains, the plasmids pcadCPcadAlux and p602/22lux were inserted into Staphylococcus aureus RN4220 [72] and Bacillus subtilis BR151 [73] or BR151(pBL1) [38] by electroporation [74, 75].

Cultivation of bacteria

The constructed strains were stored as frozen glycerol (15%) stocks at -70°C. Prior to testing, bacteria were grown on LB [26] agar plates containing appropriate antibiotics (see below). For the test, bacteria were cultivated in M9 medium (per liter: 6 g of Na₂HPO₄, 3 g of KH₂PO₄, 0.5 g of NaCl, 1 g of NH₄Cl, 0.25 g of MgSO₄ × 7H₂O, 0.01 g of CaCl₂; [26]) supplemented with glucose (final concentration 1 g l^-1) and acid hydrolysate of casein (cas-AA; LabM; final concentration 5 g l^-1), HMM medium (per liter: 8.4 g of MOPS at pH 7.2, 3.7 g of KCl, 0.53 g of NH₄Cl, 0.07 g of MgSO₄ × 7H₂O, 0.21 g of glycerol-2-phosphate, 0.0003 g of FeCl₃ × 6H₂O; all the components were prepared as 10 or 100-fold concentrated stocks, filter-sterilized and stored at +4°C except for glycerol-2-phosphate, which was stored at -20°C; [76]) supplemented with glucose (final concentration 1 g l^-1) and cas-AA (final concentration 5 g l^-1) or LB medium. The culture media were supplemented with antibiotics as follows: ampicillin 100 μg ml^-1 (E. coli MC1061(pmerR_BSBPmerlux), MC1061(pSLlux)), ampicillin 100 μg ml^-1 plus tetracycline 10 μg ml^-1 (E. coli MC106 (pSLcueR/pDNPcopAlux), MC1061(pSLzntR/pDNPzntAlux)), tetracycline 10 μg ml^-1 (MC1061(pDNlux)) and 20 μg ml^-1 (P. fluorescens OS8(pDNmerR_BSBPmerlux), OS8(pDNpbrRPpbrAlux), OS8(pDNcadRPcadAlux), OS8(pDNlux)), kanamycin 50 μg ml^-1 (S. aureus RN4220(pcadCPcadAlux), RN4220(p602/22lux) and B. subtilis BR151(pcadCPcadAlux)) or 100 μg ml^-1 (P. fluorescens OS8::KnmerR_BSBPmerlux, OS8::KncueRPcopAlux, OS8::KnzntRPzntAlux, OS8::KnpbrRPpbrAlux, OS8::KncadRPcadAlux, OS8::Knlux), kanamycin 50 μg ml^-1 plus erythromycin 10 μg ml^-1 (BR151(pBL1/p602/22lux)). All cultivations were performed on a shaker at 200 rpm, 30°C.

M9 medium with the above-mentioned supplements was used to cultivate E. coli and P. fluorescens strains when they were use to test Cd, Hg, methylmercury, Zn, Cu or Ag. HMM medium with supplements was used to cultivate E. coli and P. fluorescens if they were used to test Pb. The cultures in exponential phase (OD₆₀₀ = 0.6–0.8) were diluted with the respective test media until OD₆₀₀~0.1 except the four P. fluorescens strains with chromosomal insertions, which were cultivated until the OD₆₀₀ of 0.8 (OS8::KncadRPcadAlux and OS8::KncueRPcopAlux), 0.4 (OS8::KnmerR_BSBPmerlux) and 0.2 (OS8::KnzntRPzntAlux).

LB medium was used to cultivate S. aureus and B. subtilis strains. The cultures in exponential phase (OD₆₀₀ = 0.6–1) were then washed twice with either glucose and cas-AA supplemented M9 medium (further test with Hg, Zn or Cd) or cas-AA and glucose-supplemented HMM (further test with Pb) and diluted until OD₆₀₀~0.1. The number of bacterial cells in the test was determined for each strain before the test by spreading the appropriate dilutions of bacterial culture on antibiotic-containing LB agar plates and counting the colony forming units after 24 hours growth at 30°C.

Preparation of standard chemicals

All heavy metal salts were at least of analytical grade (98%): mercury (II) chloride (HgCl₂), methylmercury chloride (MeHgCl), cadmium (II) chloride (CdCl₂), zinc (II) sulphate-7-hydrate (ZnSO₄ × 7H₂O) and copper (II) sulphate-5-hydrate (CuSO₄ × 5H₂O) were purchased from Riedel-de-Haën, lead (II) nitrate (Pb(NO₃)₂) was from Fluka and silver (I) nitrate (AgNO₃) from J. T. Baker B.V. 1 M stock solutions of CdCl₂, ZnSO₄ × 7H₂O and CuSO₄ × 5H₂O as well as 0.1 M of HgCl₂, AgNO₃ and Pb(NO₃)₂ were prepared in MilliQ water. 0.1 M stock solution of methylmercury chloride was prepared in DMSO. Before the induction measurements, serial dilutions of the stock solutions (geometrical factor 3.3) were done in MilliQ water.

Measurements and calculations

Serial dilutions of heavy metal standard solutions in MilliQ water were pipetted onto white 96-well Cliniplate (Thermo Labsystems, Helsinki, Finland) (100 μl per well), and an equal volume of the bacterial suspension (preparation escribed above) was added and the plates were incubated for 2 h at 30°C (optimal for the induction of bioluminescence in sensor cells) without shaking. After incubation, luminescence was measured with plate luminometer Fluoroskan (Thermo Labsystems). All the measurements with sensor and respective control bacteria were performed as three independent experiments in duplicate. To determine the background luminescence (non-affected by added metals) for each sensor and control strain, MilliQ water was added to bacterial suspension instead of metal dilutions (6 replicates).

Response of either metal sensor or control bacteria was expressed as normalised luminescence (NL) calculated as in [63] according to the following formula:

(1)

where L_B is the background luminescence (luminescence of bacteria without added heavy metal) and L_S the metal-affected luminescence of the respective bacterium. The NL of the control strains could be used to correct the luminescence of the sensor stains in slightly toxic, coloured or turbid samples.

The limit of determination in NL values (LOD_NL) for the sensor bacteria was determined in every assay as follows:

(2)

where is the mean background luminescence value of the sensor (minimum six blanks included in each assay) and SD is the standard deviation. The LOD as mg of metal l^-1 was the metal concentration corespondng to LOD_NL in the calibration curve.

In the case of the control strains, the inhibition of luminescence by metals was calculated by the following formula:

(3)

The 2-h EC₅₀ (concentration of metal causing 50% decrease in luminescence during 2 hours of exposure) was calculated from concentration-effect curves using log-normal regression.