Metagenomics-Guided Discovery of Potential Bacterial Metallothionein Genes from the Soil Microbiome That Confer Cu and/or Cd Resistance

The metagenomics-guided procedure developed here bypasses the difficulties encountered in classic PCR-based approaches and led to the discovery of novel MT genes, which may be useful in developing bioremediation tools. The procedure used here expands our knowledge on the diversity of bacterial MTs in the environment and may also be applicable to identify other functional genes from eDNA.

ABSTRACT

Metallothionein (MT) genes are valuable genetic materials for developing metal bioremediation tools. Currently, a limited number of prokaryotic MTs have been experimentally identified, which necessitates the expansion of bacterial MT diversity. In this study, we conducted a metagenomics-guided analysis for the discovery of potential bacterial MT genes from the soil microbiome. More specifically, we combined resistance gene enrichment through diversity loss, metagenomic mining with a dedicated MT database, evolutionary trace analysis, DNA chemical synthesis, and functional genomic validation to identify novel MTs. Results showed that Cu stress induced a compositional change in the soil microbiome, with an enrichment of metal-resistant bacteria in soils with higher Cu concentrations. Shotgun metagenomic sequencing was performed to obtain the gene pool of environmental DNA (eDNA), which was subjected to a local BLAST search against an MT database for detecting putative MT genes. Evolutional trace analysis led to the identification of 27 potential MTs with conserved cysteine/histidine motifs different from those of known prokaryotic MTs. Following chemical synthesis of these 27 potential MT genes and heterologous expression in Escherichia coli, six of them were found to improve the hosts’ growth substantially and enhanced the hosts’ sorption of Cu, Cd, and Zn, among which MT5 led to a 13.7-fold increase in Cd accumulation. Furthermore, four of them restored Cu and/or Cd resistance in two metal-sensitive E. coli strains.

IMPORTANCE The metagenomics-guided procedure developed here bypasses the difficulties encountered in classic PCR-based approaches and led to the discovery of novel MT genes, which may be useful in developing bioremediation tools. The procedure used here expands our knowledge on the diversity of bacterial MTs in the environment and may also be applicable to identify other functional genes from eDNA.

INTRODUCTION

Metallothioneins (MTs) are low-molecular-weight and cysteine-rich proteins widely distributed in eukaryotes and bacteria (1, 2). The superior metal-binding capacity makes MTs one of the popular candidate biomolecules for developing bioremediation tools (3–6). A number of MTs from vertebrates and plants have been used for genetic engineering of bioremediation tools. For instance, expression of a mouse MT gene fused with an IgA protease β domain to anchor the MT onto the cell surface of Ralstonia eutropha CH34 led to an enhanced cadmium (Cd) immobilization activity of the engineered strain in soil and an improved growth of tobacco in pot experiments (7). Krishnaswamy and Wilson reported a 4-fold increase in nickel (Ni) bioaccumulation by an engineered Escherichia coli strain expressing a glutathione S-transferase–pea MT gene fused to the nixA gene (an Ni transporter) (8).

Since the first MT isolated by Margoshes and Valee in 1957, numerous MTs have been discovered experimentally or bioinformatically. Phylogenetic classification by early studies separated MTs into 15 families (9). Now, it is generally considered that all MTs from different families are not homologous (2). Currently, more than 290 records of MTs can be found in the UniProt database (10), with dozens of them experimentally validated, such as human MT1A (11, 12), mouse MT1 (13, 14), and Arabidopsis MT1A (15–17). Structure, biochemistry, and evolutionary relationships among the eukaryotic MTs have been extensively studied in vertebrates and plants, whereas only a limited number of MTs have been experimentally identified in bacteria to date (2).

Most of our current knowledge on bacterial MTs is based on the analysis of SmtA and MymT from Synechococcus elongatus PCC7942 (18–21) and Mycobacterium tuberculosis H37Rv (22), respectively. Meanwhile, an in vitro study led to the discovery of two bacterial MT families, the BmtA family, including MTs from Anabaena PCC7120, Pseudomonas putida KT2440, and Pseudomonas aeruginosa PAO1, and the GatA family, including MTs from Escherichia coli (23). Recently, two additional bacterial MTs were characterized experimentally, NmtA from Nostoc sp. strain PCC7120 (24) and the PflQ2 MT from Pseudomonas fluorescens Q2-87 (25). The known bacterial MTs, similar to the ones in eukaryotes, are rather divergent in sequence, despite their highly convergent function in metal binding. Although SmtA and MymT share some conserved cysteine-rich metal binding motifs (Cys motifs) with eukaryotic MTs, the preference for these Cys motifs as well as the organization of Cys motifs are distinct among kingdoms (2, 26–28). Dozens of bacterial MTs were detected by homology-based protein prediction, and a strong difference in Cys motif preference was found among the bacterial phyla (1, 2, 23), which may indicate that bacterial MTs are very diverse and that many of them are yet to be discovered with functional verification.

A major difficulty that hindered the discovery of novel bacterial MTs is the cultivation of metal-tolerant bacterial strains. As mentioned above, all identified bacterial MTs were discovered based on isolation of metal-tolerant strains. Previous studies have also explored putative MTs from sequenced bacterial genomes through the sequence analysis of Cys motifs (1, 29). However, experimental validation of the functionality of these MTs is time consuming.

Metagenomics provides a culture-independent and premise-free exploration of functional genes from environmental DNA (eDNA) (30–33). By coupling with functional genomics, metagenomics has been successfully used for detecting novel antibiotic resistance genes (34, 35), biocatalysts (36–38), and abiotic stress resistance genes (39, 40). Recently, a metatranscriptome study led to the characterization of a new metallothionein family (41). Unfortunately, for many genes, such as those encoding MTs, it remains challenging to discover novel genes from eDNA in a high-throughput manner through functional metagenomics, which is mainly due to the lack of a specific screening method.

In this study, we developed a procedure for identifying novel bacterial MTs from soil eDNA samples, by coupling metagenomic sequencing, local BLAST, evolutionary trace analysis (ETA), and chemical synthesis of the MT genes followed by functional genomics. A diversity loss experiment was carried out to enrich heavy-metal-resistant microbes in soil. A local MT database was built up for recovering full-length MT genes from the metagenome through the local BLAST procedure. ETA was then performed to evaluate the putative MT genes, after which the genes with high confidence were chemically synthesized and subjected to functional genomic analysis. The procedure used here is supposed to consistently expand our knowledge on the diversity of bacterial MTs and may also be applicable to identify other functional genes from eDNA. MTs detected in this study that conferred increased Cu/Cd resistance and biosorption may be useful for developing bioremediation tools.

RESULTS

Enrichment of metal-resistant bacteria in soil.

Soil samples were treated with different concentrations of Cu for 1 week and were subjected to amplicon sequencing. Bacterial 16S sequencing generated 15,058 to 19,317 clean reads for each soil sample (see Table S1 in the supplemental material), with a mean length of around 420 bp. A total of 1,046 to 2,959 operational taxonomic units (OTUs) were identified in each of the five soil treatments (Table 1). Chao1 and Shannon indices showed that Cu addition generally reduced bacterial diversity, with the lowest Chao1 diversity found in the soil samples treated with 8,000 mg Cu per kg of soil (Table 1; Fig. S1). Permutational multivariate analysis of variance (PERMANOVA) revealed a significant effect (r² = 0.25, P = 0.001) of Cu concentrations on bacterial community. Beta diversity analysis based on nonmetric multidimensional scaling (NMDS) showed that the five soil treatments were separated into three distinct clusters, with the stress number of 0.088: the first is the 0 mg Cu/kg, the second comprises the treatments of 1,000, 2,000, and 4,000 mg Cu/kg, and the third is the 8,000-mg Cu/kg treatment (Fig. 1).

TABLE 1.

Attributes of bacterial community diversity in Cu-contaminated soil based upon relative abundance of operational taxonomic units^a

Cu amount (mg/kg soil)	No. of OTUsb	CHAO1c	Shannon Hc
0	2,959.33 ± 68.50	573.39 ± 38.52 A	3.36 ± 0.04 A
1,000	2,103.00 ± 161.39	546.43 ± 19.72 AB	3.04 ± 0.10 B
2,000	1,858.33 ± 477.33	491.64 ± 78.71 BC	3.05 ± 0.09 B
4,000	2,231.67 ± 255.07	559.74 ± 36.73 AB	2.98 ± 0.18 B
8,000	1,046.67 ± 104.00	415.46 ± 22.03 C	3.29 ± 0.11 A

^aData are means ± standard deviations (n = 3).

^bAn OTU is defined as sequences sharing >97% similarity.

^cDifferent uppercase letters indicate significance (P < 0.05).

FIG 1.

The differences of bacterial community composition among soil samples. Ordination of microbiomes in soil was conducted by nonmetric multidimensional scaling of terminal restriction fragment length polymorphism-derived data and distance was based on Bray-Curtis dissimilarity coefficient. Treatments: 0, 1,000, 2,000, 4,000, and 8,000 represent 0, 1,000, 2,000, 4,000, and 8,000 mg Cu/kg soil, respectively.

OTU analysis showed a clear difference in bacterial community composition among the soils treated with different Cu concentrations (Fig. 2A). The top most abundant OTUs (Fig. 2B and C) were well annotated to the species level with 99% confidence. Cu addition (1,000 to 8,000 mg/kg) appeared to increase the abundance of Cupriavidus sp., Pseudoduganella eburnea, and Lysobacter panacisoli in soil, while Stenotrophomonas maltophilia, Mesorhizobium amorphae, and Serratia marcescens were particularly more abundant in the soil treated with the highest Cu concentration (8,000 mg/kg) (Fig. 2B and C).

FIG 2.

Heat map of bacterial abundance in soil. (A) Bacteria with abundance equal to or greater than 1% in the whole bacterial community. (B) Prevalent bacteria in 1,000- to 4,000-mg Cu/kg soil contaminated soil. (C) Prevalent bacteria in 8,000-mg Cu/kg soil contaminated soil. Count data were normalized by converting to log₁₀(count + 1) before being subjected to heat map generation.

Metallothionein database and relevant bioinformatic analyses.

Currently, the UniProt database contains 290 records of reviewed MTs, among which only six are from prokaryotes. Here, 84 bacterial MTs were involved to build a database for the local BLAST study. The included MTs were carefully checked for their length, species origin, Cys motifs, and, particularly, the content of Cys+His. Meanwhile, 18 MTs from human and mouse, 19 MTs from Arabidopsis and rice, and three MTs from yeast were also included for a wider comparison between bacterial MTs and eukaryotic MTs to further explore the sequence features of MTs. These features, as stated above, are used as criteria to evaluate putative MTs detected from the metagenomes.

All bacterial MTs have sequence lengths that range from 49 (Cyanothece sp. strain CCY0110) to 113 (Cystobacter fuscus DSM 2262) amino acids (aa), while those of eukaryotic MTs vary from 45 (MT1A, MT1B, and MT1C from Arabidopsis) to 87 (MT21A from rice) aa. We analyzed the statistics for both Cys and His contents. In addition to Cys, His has also been shown to play a role in the coordination of metals in metal-binding motifs (2, 42). Across-phylum comparisons showed that prokaryotic MTs contain a lower Cys+His content spanning a wider range (Fig. 3). Statistically, the included bacterial MTs contain 9 to 19 Cys+His with a median of 12, which accounts for 8.9% to 33.3% of the total residues. In comparison, human and mouse MTs contain 18 to 21 Cys+His with a median of 20, which accounts for 29.4% to 34.4% of the total residues. Interestingly, eukaryotes rarely use His for MT metal binding, which is much different from the bacterial ones. For instance, Actinobacteria MTs generally use one His in each of their Cys motifs ⁴¹CEHC⁴⁴ and ⁶¹CCAHC⁶⁵ (as in Amycolatopsis halophila YIM_93223, UniProtKB/TrEMBL protein identifier number [no.] W9DN38), while Cyanobacteria MTs have a conserved His in a ⁴⁷CGHGGCTC⁵⁴ motif (as in Synechococcus sp. no. B1XL53).

FIG 3.

Statistics of cysteine (Cys) and histidine (His) residue contents of MTs in the local database. MTs were classified into five categories, and sample size (N) is marked. Three bacterial MTs that poorly clustered in the sequence analysis were not included in these statistics. At, Arabidopsis thaliana.

Based on the alignment of homologous MTs with the experimentally verified MTs, some sequences from the database annotated as MTs were not included in the database, as they did not meet the criteria. For example, MTs from “Candidatus Dependentiae” bacterium ADurb.Bin246, Sulfurospirillum sp. strain SCADC, Arcobacter thereius, and Helicobacter pylori FD535 were excluded considering that they have a low content of Cys+His, despite the fact that they all contain Cys motif CXXC.

Alignment of the bacterial MT protein sequences showed a phylum preference of conserved Cys/His motifs, which have been used for comparison of MTs in previous reports (41, 43, 44). Cyanobacteria (group II) and Proteobacteria (group III) shared a similar pattern, which contains a ⁹CXCXXCXC¹⁶ motif at the N terminus, followed by three conserved Cys/His residues (Cys-32, Cys-36, and His-40; as in Synechococcus sp. no. B1XL53) (Fig. 4). Particularly, Cyanobacteria (group II) MTs contain a conserved His-49, and this conserved His is also detected in the relevant region in several Proteobacteria MTs (His-49, as in Anabaena sp. strain 39858, no. A0A1W5CMH6) (Fig. 4). In contrast, Actinobacteria (group I) MTs have three conserved Cys motifs: ⁴CEVC⁷, ⁴¹CEHC⁴⁴, and ⁶²CAHC⁶⁵ (as in Amycolatopsis halophila YIM_93223, no. W9DN38) (Fig. 4). Cross-domain comparisons showed that these above-mentioned phylum preferences in Cys motifs were different from those of all the investigated eukaryotic MTs. Specifically, plants, humans, and mice use several sets of standalone CXC. For example, human MT2A contains ⁵CSC⁷, ¹³CTC¹⁵, ¹⁹CKC²¹, ²⁴CKC²⁶, ³⁴CSC³⁶, ⁴⁸CIC⁵⁰, and ⁵⁵CSC⁵⁷. Five conserved CXC motifs are also found in the rice-Arabidopsis thaliana-yeast group, although MTs from here are phylogenetically diverse (data not shown).

FIG 4.

Protein sequence alignment of MTs from Actinobacteria (group I), Cyanobacteria (group II), and Proteobacteria (group III). Sequence alignments were performed using ClustalW. Conserved cysteine (C) and histidine (H) residues are highlighted in red and champagne, respectively. The designations that appear at the far left are protein identifiers from the UniProtKB/TrEMBL database. *, truncated N terminus.

Recovering MTs from soil metagenome.

Based on the species composition cluster determined by NMDS, the 4,000-mg/kg Cu treatment was chosen for shotgun metagenomic sequencing. This treatment resulted in a prominent enrichment of Cu-resistant bacterial OTUs and adequate DNA quality for shotgun metagenomic sequencing. The sequencing generated 14.39 GB of clean data, and a 330.7-Mb assembled sequence was obtained, with an N₅₀ value of 995 bp and GC content of 62.37% (see Table S2). A total of 574,449 genes were annotated.

At cutoff values of 60 bp for alignment length and an E value of 10⁻¹⁰, we obtained 90 records of candidate MTs from the local BLASTN results after duplication removal. From the metagenome, a total of 32 full-length putative MTs were recovered, among which 27 were assumed to possess metal-binding function with high confidence based on a manual evaluation of the MTs’ lengths, Cys+His contents, and Cys motifs. These 27 putative MTs were designated MT1 to MT27 (Fig. 5). Protein sequence lengths of the 27 putative MTs ranged from 52 aa (MT2) to 166 aa (MT11), and theoretical isoelectric points (pIs) varied from 4.73 (MT2) to 12.04 (MT9). Meanwhile, the 27 putative MTs contained 11.1% to 22.2% Cys+His (see Table S3). Alignment of the 27 MTs showed that most of them contained a conserved ⁴⁹CXHCXC⁵⁴ motif (as in MT1). Most of the MTs also harbored another three conserved motifs: ¹³CXXC¹⁶, ⁴⁰CXXH⁴³, and ⁶⁹CCXHC⁷³ (as in MT1) (Fig. 5). As mentioned above, CCXCC is a common metal-binding motif found in humans and mice. Notably, except for three outliers, all the MTs had two standalone conserved His residues (His-33 and His-59, as in MT1), and MT3, -13, -20, and-21 had an additional ¹²CFEC¹⁵ motif (as in MT3). We also performed sequence alignment of some bacterial MTs which were experimentally identified. These include the M. tuberculosis MymT from the phylum Actinobacteria, the S. elongatus SmtA from PCC7942 and Nostoc NmtA from the phylum Cyanobacteria, and E. coli and Pseudomonas MTs from the phylum Proteobacteria. Generally, these experimentally identified MTs use different patterns of Cys/His metal-binding motifs (Cys motif clusters and conserved His), which are all distinct from the 27 potential MTs detected in this study. Homologs of the 27 putative MTs were further searched against the UniProt database, and best match proteins were all found to be uncharacterized, with an identity of 39.8% to 90.9% (Table S3).

FIG 5.

Protein sequence alignment and Cys/His motif pattern of the 27 putative MTs identified in this study and the prokaryotic MTs experimentally verified in previous studies. Sequence alignments were performed using ClustalW. Conserved histidine residues (H) are highlighted in champagne, and conserved cysteine residues are highlighted in red. *, truncated N terminus; X/Xs, optional amino acid(s).

Screening for Cu/Cd-tolerant MTs in common E. coli cells.

The 27 putative MTs were then subjected to functional screening and identification, using different E. coli strains that are listed in Table 2. First, the 27 MT genes were chemically synthesized and expressed in E. coli strain BL21(DE3) cells using the pET-28a vector. Transformed strains were subjected to a drop assay, and the results showed that all the strains harboring MTs or empty pET-28a vector had similar growth without Cu exposure after 5 days incubation (Fig. 6A). Strains expressing MT5, -15, -20, -21, -23, or -27 grew under 4.7 mM Cu stress, which completely inhibited the growth of the control (Fig. 6). Under 4.8 mM Cu exposure, only the strains expressing MT15, -20, or -21 exhibited distinct growth (Fig. 6A). For the Cd resistance screening, all the strains grew on the LB plates without Cd exposure after 3 days (Fig. 6A); strains expressing MT6, -18, -20, or 27 showed better growth under 0.8 mM Cd stress, while MT18 and MT27 conferred a distinct growth promotion within 3 days when the host cells were exposed to 0.9 mM Cd (Fig. 6A).

TABLE 2.

E. coli strains used in this study

Name	Genotype	Source
BL21(DE3)	F− ompT hsdSB(rB− mB−) gal dcm (DE3)	Invitrogen
DH5α	F− φ80lacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rK− mK+) phoA supE44 thi-1 gyrA96 relA1 λ−	Invitrogen
RW3110	F− λ− IN(rrnD-rrnE)1 zntA1(CdS ZnS)::kan rph-1	CGSC
JW0473-3	F− Δ(araD-araB)567 ΔlacZ4787(::rrnB-3) ΔcopA767::kan λ− rph-1 Δ(rhaD-rhaB)568 hsdR514	CGSC

FIG 6.

Screening of functional MTs under Cu/Cd stress. (A) Drop assay was performed to test the Cu/Cd resistance of E. coli strains expressing the 27 MTs. The strain harboring empty vector was used as a negative control. Legend shows the position of each drop of strains in the Luria-Bertani plates. Three biological replicates were performed, and one typical observation is shown. (B) Strains transformed with six selected MTs were subject to a 12-h incubation in liquid Luria-Bertani medium supplemented with 0.6 mM Cd or 3.5 mM Cu. Control indicates strains transformed with empty pET-28a vector, and three biological replicates were performed. Cells expressing MTs were incubated in liquid Luria-Bertani medium supplied with 100 μM Cu (C) or 10 μM Cd (D) for 12 h, after which Cu/Cd/Zn/Mn concentration of each strain was measured. Values are presented as means ± standard deviations (SDs). *, P < 0.05 versus control.

To further investigate Cu/Cd resistance of MTs, six MTs (MT5, -6, -15, -18, -20, and -27) that conferred to the hosts an improved Cu and/or Cd resistance were selected and subjected to a growth curve test. Generally, all tested six MTs conferred to the host strains better growth under Cu or Cd stress (Fig. 6B). After a 12-h incubation, strains expressing MT27 or MT18 showed the maximal optical density at 600 nm (OD₆₀₀) (Fig. 6B).

Cu and Cd sorption of the strains expressing the six MTs was also determined. Results showed that under 100 μM Cu exposure, strains expressing MT18 had a significant increase in Cu accumulation. Meanwhile, all six MTs led to a strong increase in Cd accumulation for the host cells under 10 μM Cd exposure, among which a 13.7-fold increase was found for MT5 (Fig. 6C and D). Interestingly, all six MTs conferred a significant increase in Cu accumulation at 10 μM Cd. Mn and Zn bioaccumulation in the transformed strains was also measured, and the results showed that the expression of all six MTs led to a substantial increase in Zn uptake under Cu or Cd stress, while only MT5, -6, -15, and -20 significantly increased Mn uptake in the hosts under Cd stress.

Verification of functioning MTs in Cu/Cd-sensitive strains.

Six MTs tested above were also assessed in a different system. A pUC-PTR vector was constructed, and representative MTs were cloned into this vector, under the promoter of a tobacco plastid 16S rRNA gene. Two metal-sensitive E. coli strains, JW0473-3 (ΔcopA) and RW3110 (ΔzntA), were transformed with those recombinant vectors to determine whether the MTs could restore the Cu/Cd-resistant phenotype. Strain JW0473-3 showed a Cu-sensitive phenotype, while RW3110 strain was more sensitive to Cd than strain DH5α (Fig. 7A and B). When expressing MT6, -18, or -27, the transformed strains exhibited improved growth under Cu exposure, but those with MT5, -15, or -20 expression failed to show improved growth (Fig. 7C). Meanwhile, expression of MT18 and MT20 tremendously promoted the growth of host cells under Cd exposure, but MT5, -15, and -27 did not enhance the Cd resistance of host cells (Fig. 7C).

FIG 7.

Validation of Cu/Cd resistance conferred by selected MTs in Cu/Cd-sensitive strains. Resistance for Cu-sensitive strain JW0473-3 (A) and Cd-sensitive strain RW3110 (B) is shown. Strain DH5α was used as a positive control. Cells were dropped on the LB plates without any additive. The plates were incubated at 37°C for 5 days (for Cu) or 3 days (for Cd). (C) Selected MT genes were introduced into the strain JW0473-3 or RW3110, and the transformed strains were subject to a drop assay. Control indicates sensitive strains harboring empty pUC-PTR vector. All drop assay experiments were performed with three biological replicates, and one typical observation is shown.

DISCUSSION

MTs are well known as metal-binding proteins, yet our knowledge on the prokaryotic MTs has been limited until now. In this study, we identified 27 putative MT genes from eDNA with high confidence, among which four enhanced the hosts’ Cu and/or Cd resistance, based on a series of growth tests in combination with metal uptake evaluation. This goal was obtained by integrating the use of metagenomic annotation, ETA, chemical synthesis, and functional genomics. The strategy used in this study may greatly expand our knowledge on prokaryotic MT diversity in the environment.

Diversity loss is common in inhibition assays which are used for quantitatively measuring the effect of metal toxicity on soil microbes (45, 46). Both total microbial abundance and functional microbial groups (e.g., ammonia-oxidizing prokaryotes) can be significantly reduced by acute Cu toxicity at the dose used in the present studies (47, 48), which was normally accompanied by an enrichment of metal-resistant populations (49). Diversity loss with an enrichment of metal-resistant populations has also been observed for chronic metal pressure under field conditions (30). In the present study, such diversity loss by Cu was supposed to enrich Cu/Cd-resistant populations in the soil. The results showed that a significant change in the microbial community structure occurred after Cu addition (Fig. 1), among which the moderate Cu dosages (1,000, 2,000, and 4,000 mg/kg) differed from the extreme Cu dose, 8,000 mg/kg, which exerted the biggest impact on the community in terms of the OTU number and Chao1 diversity index (Table 1). Specifically, it was observed that Cupriavidus sp., the OTU with the highest overall abundance, was enriched after the Cu treatments. The 4,000-mg/kg Cu treatment resulted in a 305-fold abundance of Cupriavidus sp. relative to that in the control. It is known that the genus Cupriavidus contains metal resistance species (28), including several Cu/Cd-resistant Cupriavidus metallidurans strains (50–52) (closest to OTU0 of this study) and the heavy-metal-resistant sulfite oxidizer Cupriavidus necator (53).

Metagenomic mining of functional genes from eDNA has been a popular and fruitful strategy in recent years. A wealth of genomic features with ecological and industrial interests were predicted by early studies, but many of them lacked further functional verification, such as for antibiotic resistance genes from soil (54), uncharacterized metal resistance genes (55, 56), and novel photoreceptor genes from Sargasso Sea (57). By coupling with tools of functional genomics, metagenomic mining is even more powerful for the discovery of novel functional genes. For example, antibiotic stress screening of a soil metagenomic DNA library revealed 41 novel antibiotic resistance genes from eight families (58). Yun et al. identified an amylolytic enzyme from a soil metagenomic DNA library, using a visual screening method based on enzyme activity (59). From the extreme environment, Morgante et al. explored the diversity of arsenic resistance genes and identified some unknown genes that conferred arsenic resistance (60). Our previous study also identified several Cd resistance genes from a small-size insert metagenomic library, and some of the positive clones were not directly related to heavy metal resistance according to the previous studies (61). Nonetheless, such a functional metagenomics scheme largely relies on the availability of screening methods, which may vary with research purpose and target genes (62–64). Function-based approaches basically allow the detection of genes that are convergent in a specific function but not necessarily phylogenetically associated. Therefore, conventional functional metagenomics is still limited in identification of a specific gene family.

Instead of recovering sequences from eDNA through conventional PCR, we obtained the MT DNA molecules by chemical synthesis, which is independent of PCR that possesses the problems associated with DNA quality and target gene abundance. In addition, chemical synthesis can be performed in bulk based on the bioinformatic information. Here, we combined the metagenomics annotation with a personalized MT database, ETA, chemical synthesis, and functional genomic verification. This procedure overcomes the limitation of PCR cloning and the lack of functional screening markers for MTs and therefore provides a new possibility for the exploration of a family of functional genes such as MTs.

Local BLAST with a personalized database largely promoted the annotation of the valuable genes in metagenomics data (30). Different from the previously established local BLAST procedure, an ETA was carried out for the in-depth definition of bacterial MTs. It assisted in the evaluation of putative MTs from the local BLAST search against the metagenome. Similar to existing results, a vast divergence of candidates among these known MTs from eukaryotes and prokaryotes was found by using ETA, which means that all MTs have a polyphyletic origin and cannot be deemed as a single homologous family. Meanwhile, we did not find any universal pattern that uses the same Cys motif across all the MTs. Rather, a universal pattern was mostly obvious at the phylum level (Fig. 4).

The local BLAST procedure is supposed to annotate genes homologous to those in the database. Surprisingly, the putative MT genes detected in this study were different in protein sequence from the ones in the personalized database. Sequence analysis indicated that they were all novel clones and similar to the uncharacterized homologs in the UniProt database (Fig. 5; see also Table S3 in the supplemental material). A further look showed that the conserved Cys motifs and His profiles of the 27 putative MTs were distinct from those of the consensus of group I/II/III in our local BLAST database, and this may be due to the nature of MTs, which are convergent in function but divergent in sequence.

Sequence alignment indicated that these phylogenetically unrelated MTs (except for MT9) were surprisingly conserved in the Cys motif, and this was also true for bacterial His sites (Fig. 4 and 5). Cys motifs are critical traits to define MTs (2, 29, 41), and the results obtained in the present study probably indicate that this local BLAST method is able to detect novel MTs that are phylogenetically distinct from the known MTs with conserved Cys and His sites. Coupled with sequence alignment, it further verified that these MTs were not random sequences or system errors but have unassigned homologs in the sequenced bacterial genomes (Fig. 5). Except for MT9, the Cys and His pattern of the potential MTs identified in this study is different from that of the group I/II/III MTs in our local BLAST database. Considering that these 26 MTs show protein sequence similarity with MTs of group I, sharing 12 conserved amino acids (see Fig. S2), the potential MTs identified in this study may be of bacterial origin and probably belong to the group I MT family.

Functional assays revealed that several putative MTs conferred to the host an improved Cu/Cd resistance. A growth curve monitoring assay was further performed to test the resistance capacity of the representative MT clones (Fig. 6). Considering that copA (gene identifier [ID] ECD_00435, a Cu resistance gene) and zntA (gene ID ECD_03318, a Zn/Cd resistance gene) located in the genome of host strain BL21(DE3) might affect the evaluation of cloned MTs for Cu or Cd resistance, mutant strains sensitive to Cu/Cd were used to provide an enhanced test of the selected MTs. An expression vector was purposefully designed in the present study for the heterologous expression. Expression of MT6, -18, -20, and -27 restored the Cu- and/or Cd-resistant phenotype in mutant strains (Fig. 7), which means these MTs conferred enhanced resistance even when the hosts are deficient in a heavy metal efflux system. Moreover, expression of the representative six MTs apparently changed the Cu/Cd accumulation in the host (Fig. 6), which strongly indicated the in vivo biological function of these MTs. Overall, evidences in this study highlighted the excellent Cu and/or Cd resistance capacity of MT5, -6, -15, -18, -20, and -27. Genetic evidence in this study may support the fact that these chemosynthetic MT genes obtained from eDNA were functioning in heavy metal resistance in environmental microorganisms. It should also be noted that the putative MTs that failed to confer to the host Cu/Cd resistance might have other potential biological functions. Previous discussions on host cells indicated that only approximately 40% of genes from a metagenome are expressible in E. coli (65), due to discrepancies in the gene expression systems between the original bacterium and E. coli. Codon usage optimization of environmental genes or the use of alternative hosts with an adaptive expression vector may improve the functional screening of metagenome-derived genes.

MATERIALS AND METHODS

Soil, bacterial strains, and culture conditions.

Soil used in this study was sampled from a research station located at Luancheng, Shijiazhuang (37°53′N, 114°41′E), in January 2016. The soil is a typical fluvisol with a pH of 8.12. Total organic matter and total nitrogen content of the soil were 2.12% and 0.09%, respectively.

E. coli strains [BL21(DE3), DH5α, RW3110, and JW0473-3] were used for functional genomic tests (Table 2). All strains were cultured in Luria-Bertani (LB) medium at 37°C, with ampicillin (100 mg liter⁻¹), kanamycin (50 mg liter⁻¹), or isopropyl-β-d-thiogalactoside (IPTG; 200 μM) (Sigma-Aldrich, St. Louis, MO, USA) added for different purposes. For the screening of functional genes (drop assay), the LB medium was supplied with 1.5 g liter⁻¹ agar and different concentrations of Cu (as CuCl₂·2H₂O) or Cd (as CdCl₂·2.5H₂O). For the Cu/Cd bioaccumulation assay, transformed E. coli strain BL21(DE3) (optical density at 600 nm [OD₆₀₀] = 0.2) was incubated for 24 h at 28°C in 50 ml liquid LB medium with 3.5 mM Cu or with 1.0 mM Cd.

Soil microbial diversity loss and enrichment of metal resistance microbes.

Sieved dry soil was weighed into 50-ml Falcon tubes (15 g each), treated with different concentrations of Cu (as CuSO₄·5H₂O) solution to 30% (vol/wt) water content and incubated at 25°C in the dark for 1 week. Soil moisture was monitored and adjusted using distilled water during the incubation. The concentrations of 0, 1,000, 2,000, 4,000, or 8,000 mg Cu/kg soil were used. Three replicates were conducted for the Cu inhibition incubation.

DNA extraction.

Soil was sampled at the 7th day of incubation. Soil in the tube was gently mixed before sampling by using a spoon. Soil DNA was extracted using a PowerSoil DNA isolation kit (Mo Bio Laboratories, Inc., Carlsbad, CA, USA), according to a previous study (66). DNA quality was checked by using a NanoDrop (Thermo Fisher Scientific, Waltham, MA, USA) and gel electrophoresis before being subjected to molecular experiments.

Amplicon and shotgun metagenomic sequencing.

Soil DNA concentration was measured using a Qubit 2.0 fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Bacterial 16S sequencing was performed by using the Illumina MiSeq platform. The 16S V3 to V4 region was PCR amplified using the primer pair 341F-CCCTACACGACGCTCTTCCGATCTGCCTACGGGNGGCWGCAG (barcode underlined) and 805R-GACTGGAGTTCCTTGGCACCCGAGAATTCCAGACTACHVGGGTATCTAATCC. Library preparation was performed by using the Nextera XT DNA Library Preparation kit (Illumina, San Diego, CA, USA) according to the standard manufacturer’s protocol. The library pool was sequenced on the Illumina MiSeq with paired-end 300-bp reads.

For shotgun metagenomic sequencing, purified DNA products were sheared with an S220 focused-ultrasonicator (Covaris, Woburn, MA, USA). DNA fragments were selected using Agencourt AMPure XP (Beckman Coulter, Pasadena, CA, USA). The sequencing library was prepared using the NEBNext kit (Illumina, San Diego, CA, USA). Shotgun sequencing was performed by using the HiSeq 2500 with PE150.

Metagenomic data analysis.

For 16S sequencing data analysis, paired-end reads were merged using pear v0.9.6 (67), followed by quality control through PRINSEQ v0.20.4 (68). Chimera removal and OTU clustering were performed in USEARCH using relevant commands (69). The alpha diversity was determined using Explicet software (70), with minimum library size as the cutoff size and 1,000-bootstrap resampling. Permutational multivariate analysis of variance (PERMANOVA) (71) using the vegan package (https://cran.r-project.org/web/packages/vegan/index.html) was performed on Bray-Curtis distances among bacterial communities. Beta diversity was assessed in PAST software package ver 3.16 (72), using ordination methods of nonmetric multidimensional scaling (NMDS). Statistical analyses were conducted using SAS (version 9.4; SAS Institute, Cary, NC). Data were subjected to analysis of variance (ANOVA) and means separation using Fisher’s least significant difference (LSD) test, with a P value of ≤0.05 considered significant. Functional prediction from the 16S sequencing analysis results was performed using PICRUSt v1.0.0 (73). Representative OTUs were selected for phylogenetic analysis with MEGA 7 (74).

For shotgun metagenomic data analysis, raw sequence trimming and quality control were performed using Trimmomatic (75). Clean reads were further assembled by IDBA_UD (76). Contigs were then subjected to a local BLASTX search for candidate microbial MT genes.

MT database for local BLAST.

Prokaryotic MT protein sequences were gathered by searching the UniProt database using “metallothionein” as an entry. An MT protein database was then built for subsequent local BLAST searches. The database contained 84 sequences, with 36, 28, 19, and 1 sequence from Cyanobacteria, Proteobacteria, Actinobacteria, and a candidate phylum, respectively. Known and reviewed eukaryotic MTs from human, mouse, rice, yeast, and Arabidopsis thaliana were also retrieved for comparisons. All the MTs were aligned in ClustalW (77). Sequence statistics of the known MTs were conducted with the Sequence Manipulation Suite (78), and (i) the protein sequence length, (ii) cysteine/histidine residue (Cys/His [C/H]) content, and (iii) metal-binding motifs were used as criteria to evaluate the candidate MTs from the local BLAST results. Meanwhile, considering that most of the prokaryotic MTs from UniProt are annotated based on homologous alignment, the MTs included in the database for local BLAST were carefully checked as well, based on the criteria.

Local BLAST search for MTs.

Local BLAST search of the metagenome against the MT database was applied according to the procedure built in our previous studies (30, 55). The cutoff expectation E value was set as 10⁻¹⁰. The BLAST results were imported into an Excel worksheet and sorted based on aligned length. A threshold of 30 bp was used for screening putative MT genes in the metagenomics data set. All the contigs containing candidate MTs were picked out manually and then subjected to gene finding using Glimmer (79) and ORFfinder (https://www.ncbi.nlm.nih.gov/orffinder/). The selected contigs were aligned as well against the GenBank database one by one, and the sequence regions annotated as MTs were compared with the open reading frame (ORF) found by Glimmer and ORFfinder. The candidate ORFs located around the regions aligned in the local BLAST results were picked out in DNAMAN (Lynnon Biosoft, San Ramon, CA, USA) and then manually curated against the GenBank database. The sequences with highest similarity in GenBank were retrieved for phylogenetic analysis.

Synthesis and recombination of MT genes.

Predicted MT genes obtained from the soil metagenome sequences were chemically synthesized (GENEWIZ, Suzhou, China). Cleavage sites of restriction enzymes NdeI and EcoRI were added onto the 5′ end and 3′ end of MT DNA fragments, respectively. The chemically synthesized MTs with cleavage sites were then doubly digested using NdeI and EcoRI restriction enzymes (New England BioLabs, Ipswich, MA, USA), followed by separating and retrieving target fragments via agarose gel electrophoresis. The MT fragments were then introduced into pET-28a vector with NdeI and EcoRI sites using T4 DNA ligase (New England BioLabs, Ipswich, MA, USA). The recombinant vectors were doubly digested with NdeI and EcoRI restriction enzymes to examine whether the insertions were successful and were double checked by Sanger sequencing.

Functional screening of MTs contributing to Cu/Cd resistance.

The MTs with Cu or Cd resistance were screened via a drop assay described in our previous study (61). In brief, recombinant pET-28a vector harboring MT genes was first introduced into E. coli BL21(DE3) cells by using a CaCl₂-based chemical transformation method; the transformed cells were then incubated overnight, harvested by centrifugation, and resuspended in water (OD₆₀₀ = 1.0). A total of 5 μl cell solution was spread onto the LB plates containing 50 mg liter⁻¹ kanamycin, 200 μM IPTG, and different concentrations of Cu or Cd. For Cu, 0, 4.7, and 4.8 mM was applied for the drop assay; for Cd, 0, 0.8, and 0.9 mM was applied. The metal concentrations were chosen based on previous experiment results. The plates were incubated at 28°C for 5 days (for Cu) or 3 days (for Cd), and the growth status of the colonies was evaluated. The strain harboring empty pET-28a vector was used as a control.

Growth curve.

The six candidate MTs with Cu/Cd resistance were further examined by performing a growth test of transformed E. coli strains. Briefly, each strain was inoculated in 50 ml LB liquid medium containing 3.5 mM Cu or 0.6 mM Cd. The metal concentrations in the liquid culture were chosen based on MIC tests. Kanamycin (100 g liter⁻¹) and IPTG (200 μM) were also supplied in the medium to maintain the vector and to induce the expression of MT genes, respectively. The strains were cultured with a starting OD₆₀₀ of 0.2 at 28°C at 165 rpm for 8 h. OD₆₀₀ values were measured every hour by using a spectrophotometer (Eppendorf, Hamburg, Germany). The strain harboring empty pET-28a vector was used as a control.

Expression of MTs in E. coli ΔcopA or ΔzntA strain.

Cu or Cd resistance associated with the selected six MTs was evaluated by expressing them in the Cu-sensitive strain JW0473-3 (ΔcopA) and Cd-sensitive strain RW3110 (ΔzntA). A novel vector was constructed for MT recombination. A tobacco plastid 16S rRNA gene promoter (Prrn), a heterologous transcriptional enhancer of bacteriophage T7 gene 10 (T7g10), a multiple cloning site, and a rrnB T1 terminator were chemically synthesized and introduced into backbone vector pUC19 with HindIII and EcoRI restriction enzyme sites to generate a pUC-Prrn-T7g10-rrnB1 (pUC-PTR) vector. Prrn was reported as a constitutive promoter (80), T7g10 was shown as an enhancer which can promote translation of the target gene (80, 81), and rrnB T1 is commonly used and identified as a strong terminator (82). MT genes were amplified from the above-mentioned recombinant pET-28a by PCR and introduced into pUC-PTR vector. The recombinant vectors were transformed into E. coli strain RW3110 or JW0473-3 via CaCl₂-based chemical transformation. A drop assay was performed to examine the effect of MT recombination as described above with minor modifications. The transformed strains were collected and suspended in water (OD₆₀₀ = 1.0), followed by serial dilution. Ampicillin (100 mg liter⁻¹) was supplied in the LB medium as required; 3.5 mM Cu or 0.35 mM Cd was applied to the medium. The metal concentrations in the solid culture were chosen based on MIC tests. The plates were incubated at 37°C for 48 h, and the growth of the colonies was observed. The strain DH5α was used as a positive control, and the strain harboring empty pUC-PTR vector was used as a negative control.

Cu/Cd bioaccumulation.

The cells expressing MT genes were incubated for 12 h as described previously and harvested by centrifugation at 4,000 × g. The cells were then washed with NaCl solution (3% [wt/vol]) and then oven dried and digested with 5 ml of HNO₃ (61). The digested solution was then diluted with Millipore water and subjected to metal determination using a ZEEnit 700 P atomic absorption spectrometer (Analytik Jena, Jena, Germany) equipped with a flame atomizer. Certified reference material laver (GWB10023, certified by IGGE) was used as a standard reference.

Data availability.

All high-throughput sequencing data have been deposited into GenBank under accession numbers SRR11006123 to SRR11006137 (SRA amplicon sequencing), SRR11031422 (SRA metagenomic sequencing), PRJNA604302 (BioProject), and SAMN13964729 to SAMN13964733 (BioSample). The nucleotide sequences of the 27 MT genes can be found in GenBank under accession numbers MT035804 to MT035830.