Diversity of the Tellurite Resistance Gene Operon in Escherichia coli

Tellurite is highly toxic to most bacteria owing to its strong oxidative ability. However, some bacteria demonstrate tellurite resistance. In particular, some Escherichia coli strains, including Shiga toxin-producing E. coli O157:H7, are known to be resistant to tellurite. This resistance is involved in ter operon, which is usually located on a prophage-like element of the chromosome. The characteristics of the ter operon have been investigated mainly by genome analysis of pathogenic E. coli; however, the distribution and structural characteristics of the ter operon in other E. coli are almost unknown. To clarify these points, we examined 106 E. coli strains carrying the ter operon from various animals. The draft genomes of 34 representative strains revealed that ter operons were clearly classified into four subtypes, ter-type 1–4, at the nucleotide sequence level. Complete genomic sequences revealed that operons belonging to three ter-types (1, 3, and 4) were located on the prophage-like elements on the chromosome, whereas the ter-type 2 operon was located on the IncHI2 plasmid. The positions of the tRNASer, tRNAMet, and tRNAPhe indicated the insertion sites of elements carrying the ter operons. Using the PCR method developed in this study, 106 strains were classified as type 1 (n = 66), 2 (n = 13), 3 (n = 8), and 4 (n = 17), and two strains carried both types 1 and 2. Furthermore, significant differences in the minimum inhibitory concentration (MIC) of tellurite were observed between strains carrying ter-type 4 and the others (p < 0.05). The ter-type was also closely related to the isolation source, with types 2 and 4 associated with chickens and deer, respectively. This study provided new insights related not only to genetic characteristics of the ter operons, but also to phenotypic and ecological characteristics that may be related to the diversity of the operon.

STECs belonging to the seven major serotypes described above are known to carry the ter operon. For example, the STEC O157:H7 Sakai strain carried the ter operon on a prophage-like 86-kb element, named SpLE1 on the chromosome (Hayashi et al., 2001;Ohnishi et al., 2002). SpLE1 also contains genes encoding integrase, urease (ure), and iron-regulated gene A (IrgA)homolog adhesin (iha) (Hayashi et al., 2001). Comparative genome analyses indicated that the ter operons and their carrier elements are highly conserved among STEC strains belonging to the major serotypes (Ogura et al., 2009;Lorenz et al., 2017). Some STEC isolates, such as STEC O104:H4 and O178:H19, and non-STEC E. coli isolates also carry ter operons and/or show significant tellurite resistance (Orth et al., 2007;Gouali et al., 2013;Miko et al., 2014;Ferdous et al., 2015;Verhaegen et al., 2015;Kalule et al., 2018;Nguyen et al., 2021). A uropathogenic E. coli isolate was found to carry the ter operon on a large conjugative plasmid (Soltys et al., 2018). These reports suggest that the ter operon is widely distributed in E. coli including intestinal and extraintestinal pathogenic E. coli, as well as non-pathogenic E. coli. However, except for some STECs, the genetic characteristics of ter operons and their carrier elements have not been investigated in detail. In this study, in addition to STECs belonging to seven major serotypes, detailed analysis of ter operons and their carrier elements were performed in E. coli isolates belonging to more than 70 serotypes isolated from humans, cattle, pigs, chickens, and deer.

E. coli Strains
In total, 106 terA-positive E. coli strains isolated in Japan were used in this study (Supplementary Table 3). Humanderived STECs were isolated from patients and asymptomatic carriers through routine laboratory inspection. Twenty-four cattle-derived STEC strains, isolated in our previous study, were also used (Nguyen et al., 2021). Other E. coli strains from cattle, pigs, and deer were isolated from the rectal feces of slaughtered healthy animals. Feces were pre-cultured in modified E. coli broth (mEC) (Nissui Pharmaceutical Co., Ltd., Tokyo, Japan) at 37 • C for 18-20 h and then inoculated on deoxycholate hydrogen sulfide lactose (DHL) (Nissui Pharmaceutical Co., Ltd., Tokyo, Japan) or CHROMagar STEC (CHROMagar TM Paris, France) agar plates. E. coli strains from chickens were isolated from colibacillosis infected sites (such as under the skin and organs) of chickens on farms and slaughterhouses. The swabs were directly inoculated onto DHL agar plates. After incubation at 37 • C for 18-20 h, E. coli-like single colonies were picked from each plate. DNA for PCR was purified using the Wizard R Genomic DNA Purification Kit (Promega, Madison, WI, United States). E. coli species were confirmed by PCR using E. coli-specific gyrB sequences (Iguchi et al., 2015). The presence of the ter operon was screened by PCR using a universal primer pair (Supplementary Table 4), designed based on the terA sequences collected from the DNA database (GenBank/ENA/DDBJ) (Supplementary Table 1). The thermal conditions for PCR included 25 cycles of 94 • C for 20 s, 58 • C for 20 s, and 72 • C for 30 s. PCR was conducted using Kapa Taq DNA polymerase (Kapa Biosystems, Woburn, MA, United States) or Dream Taq DNA polymerase (Thermo Fisher Scientific, Inc.) following the manufacturer's instructions. The PCR products were electrophoresed on 2% agarose gels in 0.5 × Tris-borate-EDTA (TBE) (25 mM Tris-borate and 0.5 mM EDTA) followed by staining with ethidium bromide (1 mg/mL), and imaging under UV light.

DNA-Based O and H Serotyping
DNA-based E. coli O and H serotyping (Og-and Hg-typing) were conducted as described previously (Iguchi et al., 2015;Banjo et al., 2018;Iguchi et al., 2020). Briefly, 25 multiplex PCR kits (MP-1 to MP-25) containing 195 primer pairs targeting O-antigen encoding genes and 10 multiplex PCR kits (MP-A to MP-J) containing 52 primer pairs targeting flagellin (Hantigen) encoding genes were used for broad Og-and Hg-typing, respectively. If no PCR product was obtained using any of the multiplex PCR kits, the sample was determined as OgUT (Og untypeable) or HgUT (Hg untypeable).

Prevalence of Virulence Related Genes
The virulence profiles of stx1, stx2, and eae were screened using PCR. The primers and PCR conditions for stx1, stx2, and eae were previously described by Cebula et al. (1995) and Nguyen et al. (2021), respectively.

Genome Sequencing and Assembly
DNA for genome sequencing was extracted and purified using the Qiagen Genomic DNA Preparation Kit (Qiagen) or Maxwell R DNA Purification Kit (Promega). Illumina short-read libraries were prepared using 100 ng of extracted DNA using the Nextera DNA Library Prep Kit, and paired-end reads were generated using the MiSeq Reagent Kit (v3-600) on MiSeq (Illumina). MinION long-read libraries were prepared from 300 ng of genomic DNA using the Rapid Barcoding kit (SQK-RBK004) and were sequenced using R9 flow cells (FLO-MIN106) on MinION (Oxford Nanopore Technologies) to produce > 100 Mb of data per strain. The Illumina data were preprocessed using Trimmomatic (v0.36) to remove the adapter and lowquality sequences (Bolger et al., 2014). The quality of the Nanopore sequencing data was assessed using NanoPack (De Coster et al., 2018) a hybrid assembly with the Illumina and Nanopore reads was performed using Unicycler (v0.4.6) with the default parameters (Wick et al., 2017). The sequence data have been deposited in NCBI under the accession number, BioProject PRJDB10561.

Genome Analysis
Genome sequences were annotated using the online tool DFAST (Tanizawa et al., 2017) and were then visualized using an in silico Molecular Cloning software (in silico biology, Inc.). Plasmids were identified using PlasmidFinder 2.1 (Carattoli et al., 2014). The family of transposase genes was identified using the online tool ISfinder (Siguier et al., 2006). DNA sequences were aligned using the ClustalW program provided by DNA Data Bank of Japan 1 . A phylogenetic tree was constructed using the neighbor-joining method with 1000 bootstrap replicates in MEGAX (Stecher et al., 2020). Genetic comparison and dot plot analysis of the elements was conducted using GenomeMatcher 3.14 (Ohtsubo et al., 2008).

Subtyping PCR for the ter Operon
The PCR primer sequences used for ter operon subtyping are shown in Supplementary Table 4. The thermal conditions for PCR were the same as those of PCR using the universal primers targeting terA.

Multilocus Sequence Typing
Multilocus sequence typing (MLST) was performed using the internal sequences of seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, and recA) (Wirth et al., 2006). The primers used were described in a previous study (Nguyen et al., 2021). Sequence types (STs) were determined according to the 1 https://clustalw.ddbj.nig.ac.jp FIGURE 1 | Structure of the ter operon (A). Phylogenetic tree of the ter operon (B). The tree was created based on the ter operon sequences (4,546-4,551 bp) from 34 Escherichia coli strains used in this study, and 27 E. coli genomes registered in DNA database. Four of them were isolated in Japan and the remaining 23 were isolated outside Japan. Strains isolated in Japan and outside Japan are indicated by circles and squares, respectively. Open and fulfilled circles or squares indicate that the ter operon is located on the chromosome and on the plasmid, respectively. Those whose location is unknown are shown in gray. DNA sequences obtained from the DNA database (GenBank/ENA/DDBJ) are shown in blue color. Bootstrap values are displayed at the branching points.
Enterobase MLST website 2 . Concatenated sequences (3,423 bp) of the seven genes were used to construct a phylogenetic tree.

Minimum Inhibitory Concentration (MIC) of Tellurite
Tellurite resistance was evaluated on Mueller-Hinton agar plates containing twofold serial dilutions of potassium tellurite  (K 2 TeO 3 ) ranging from 0.125 to 512 µg/mL, according to the general steps of the agar dilution method (Clinical and Laboratory Standards Institutes [CLSI], 2012). Three independent experiments and two replicates were performed for each experiment. For resistant strains, the median log 2 transformed MIC values of six replicates were used for statistical analysis. The difference in log 2 transformed MIC was analyzed by Tukey's t-test and statistical significance was set at p ≤ 0.05.

Subtypes of ter Operon
First, we determined the draft genomes of 34 representative strains using the MiSeq system, which were randomly selected after sorting out so that the Og:Hg types did not overlap (Supplementary Table 3) and extracted all ter operons (4,546 − 4,551 bp) containing six genes terZ, terA, terB, terC, terD, and terE ( Figure 1A). When a phylogenetic tree was created using the ter operon sequences, it became clear that four groups were formed, named ter-type 1-4 ( Figure 1B). In particular, the nucleotide sequences of types 1, 3, and 4 were highly conserved within their respective groups. The phylogenetic trees using the sequences of each gene of 34 strains, terZ (582 bp), terA (1,158 bp), terB (456 bp), terC (1,041 bp), terD (579 bp), and terE (576 bp) showed that type-3 and type-4 formed independent groups for all genes, whereas type-1 and type-2 roughly formed their respective groups for terZ, terA, terD, and terE (Supplementary Figure 1). When a phylogenetic tree was created by adding 13 types of ter operons possessed by bacteria other than E. coli (Supplementary Table 2), three operons from Yersinia pseudotuberculosis, Salmonella enterica, and Klebsiella pneumoniae belonged to type-2, operons from seven Enterobacteriaceae genera formed an additional clade inside the group of E. coli-derived operons but no E. coli strain belonged to this clade, and the remaining three from Proteus mirabilis, Yersinia pestis, and Pseudomonas aeruginosa were distant from the E. coli-derived group (Supplementary Figure 2).

Comparison of Elements Carrying ter Operons
To analyze the structure of the ter operon as well as the surrounding region in detail, complete or almost complete genome sequences were obtained from the short-and long-read sequences of the 26 strains (Supplementary Table 3). When the attributes of ter operons were confirmed from the surrounding sequence information, the operons classified as ter-type 1, 3, and 4 were located on the chromosomes, whereas all operons classified as ter-type 2 were located on plasmids (Figure 1 and  Supplementary Figure 2).
The operon of ter-type 2 from Og136:Hg16, Og183:Hg18, Og161:Hg4, Og183:Hg42, and Og84:Hg2 was located on IncHI2 plasmids, including the conjugative system (tra genes) and some antibiotic resistant genes [aadA, aph (3'), bla, ble, cml, eptA, and tetA], with sizes ranging from 165,509 to 249,877 bp ( Figure 5  and Supplementary Figure 3). Except for the central regions downstream of the ter operons, where multiple recombination and inversions were confirmed, these plasmids roughly retained homologous gene structures. Based on the coverage of genomic sequences, the copy number of plasmids carrying the ter-operons was approximately the same as that of the chromosomes (1 vs. 0.85-1.74).
All ter operons used in this study, which were extracted from the genomes of strains other than E. coli, were located on plasmids, except for a Salmonella-derived ter-type 2 operon, and Proteus-and Yersinia-derived operons, which are on the chromosome (Supplementary Figure 2).

The Insertion Sites of Integrating Elements Carrying the ter Operon on the Chromosome
The insertion sites of elements carrying the ter operon are summarized in Figure 6. tRNA SerX and tRNA PheV were the only insertion sites of elements carrying ter-type 1 and 4 operons, respectively. Two positions of tRNA Met (ypjC-ygaQ and mug-yqhH) were the insertion sites for the elements of the ter-type 1 and 3 operons. Although the types of elements carrying the tertype 1 and 3 operons were different, a homology (98% amino acid sequence) was confirmed in the integrase genes located on the elements inserted in tRNA Met (Supplementary Figure 4). Similarly, the nucleotide sequences of the integrase in the element carrying ter-type 4 (Og54:Hg45 and Og22:Hg16 strain) were highly homologous (upper 97% amino acid sequence) with the integrases in the other integrating elements or the locus of enterocyte effacement that was inserted in the tRNA Phe positions in seven major STEC genomes (Supplementary Table 1) (data not presented). In addition, the element carrying the ter-type 1 operon of Og177:Hg25 was inserted into tRNA SerW , the element carrying the ter-type 3 operon of Og9:Hg19 was inserted into FIGURE 6 | Insertion sites of elements carrying the ter operon in the chromosome. The sites are based on the positions of tRNA in the genome of the K-12 MG1655 strain. Black, blue, and purple bars indicate elements carrying ter-type 1, 3, and 4, respectively. The width of the bars is proportionate with the element size. Hash key indicates the absence of tRNA at the site.
tRNA PheU , and the element carrying the ter-type 4 operon of Og54:Hg45 was in tRNA PheU . Although tRNA was absent in the insertion site of the ter operon-carrying elements in the Og5:Hg9 and Og109:Hg10 strains, a corresponding region surrounding the indicated tRNA was present in each strain (Supplementary Figure 5). There was an element other than the ter operon carrier inserted in the tRNA PheU position of the Og9:Hg19 strain (Supplementary Figure 5).

PCR for ter Operon Subtyping
Based on the sequences of the ter operons belonging to four subtypes, we designed type-specific PCR primers, and confirmed that these primers could specifically determine each type (Supplementary Table 4 and Supplementary Figure 6).

Distribution of ter-Types in Escherichia coli Linages
The phylogenetic tree of 80 terA-positive E. coli strains (one representative strain of the same O:H genotype and the same ter-type was selected from 106 strains) is shown in Figure 7. Overall, four ter-types were distributed in various lineages of E. coli including A, B1, B2, E, D, A × B1, and ABD FIGURE 7 | Phylogenetic tree of 80 terA-positive Escherichia coli strains. The tree was built based on the internal concatenated sequences of seven housekeeping genes (3,423 bp). The asterisk indicates eae-positive strains. Table 3), although some strains with the same ter-type formed several small clades.

Tellurite Resistance
The 106 ter-positive E. coli strains showed MICs ranging from 16 to 256 µg/mL of potassium tellurite (Supplementary Table 5), except for a ter-type 1 strain (KAP39, Og98:Hg4), which was tellurite-sensitive. The genome sequence of this strain had an IS5 inserted on the terZ due to which the gene was disrupted (Supplementary Figure 7). Based on the ter subtype, seventyfive percent of the strains were classified as ter-type 1, 2, and 3 were resistant to potassium tellurite concentrations of 64-128 µg/mL or higher, whereas only one strain (6%) in ter-type 4 was resistant to 64-128 µg/mL (Supplementary Table 5). There was no significant difference in the resistance among ter-type 1, 2, and 3 groups, whereas the resistance activity of the ter-type 4 group was significantly lower than that of the other three types (p < 0.05) (Figure 8).

DISCUSSION
The gene composition and sequence homology of the ter operon carrier elements on the chromosomes were roughly conserved for each of the ter-types, 1, 3, and 4 (Supplementary Figure 3), and strains belonging to different phylogenetic lineages carried the same types of elements (Figure 7), suggesting that at least three different types of chromosomal elements carrying the ter operons have spread horizontally within E. coli species. Six tRNA genes on the chromosome were confirmed as insertion sites for these elements, and were classified into three tRNA types: tRNA Ser , tRNA Met , and tRNA Phe . The integrase genes on the elements carrying the ter-type 1 operon were divided into two types according to the sequence homology: elements with one type of integrase gene were inserted into tRNA Ser (as seen in strains belonging to O157:H7, and so on) and elements with the other type of integrase gene were inserted into tRNA Met (as seen in strains belonging to O103:H25, and so on) (Figures 2, 6). The latter integrase gene showed high homology with that of ter-type 3 operon-carrying elements (Supplementary Figure 4), and all elements with the ter-type 3 operon were also inserted into tRNA Met , except for one strain (A150083 of Og9:Hg19) (Figure 6). Although the integrase gene was found only in two of the five ter-type 4 operon carrier elements used in this study, these two were conserved and highly homologous with the integrase in other elements of other strains that were all inserted into tRNA Phe . These results suggest that the insertion site of these elements on the chromosome depends strongly on the integrase genes and not on the basic structure of the elements. Transposases such as IS66 and IS256 were located upstream and/or downstream of the ter operons, and especially in some ter-type 1 operons, they coincided with the recombination and inversion junction points, suggesting that IS transposase was involved in the transfer of small fragments, including ter operons, to other elements.
The ter-type 2 operons were located on the IncHI2 plasmids of E. coli strains, and were also found on the IncHI2 plasmids carried by Y. pseudotuberculosis and K. pneumoniae (Supplementary Figure 2), suggesting that this type of operon was horizontally transferred in E. coli as well as within the Enterobacteriaceae through plasmids. The conjugative IncHI2 plasmid is widely found in Enterobacteriaceae and is known to mediate various drug resistance genes (Herrera-León et al., 2011;Roberts et al., 2020). In fact, the plasmids found to harbor the ter-type 2 operons in this study carried three or more drug resistance genes (Figure 5). In this study, most of the ter-type 2 operons were found in isolates from chickens (47%, 7/15). In recent years, large amounts of antibacterial drugs have been used in broiler production, and the resulting emergence and spread of drug-resistant bacteria has become a global public health concern (Nhung et al., 2017;Mehdi et al., 2018;Yang et al., 2019). The IncHI2 plasmid carrying drug resistance genes may have thus been selected in the gastrointestinal tract of chickens after many years of antibacterial drug use, and the ter operon coexisting on the plasmid was co-selected simultaneously.
All E. coli strains used in this study were isolated in Japan. However, the sequences of the ter operons obtained from the DNA database used in Figure 1 originated from E. coli isolated outside Japan (14 strains belonging to type 1, 4 strains to type 2, 3 strains to type 3, and 2 strains to type 4), except for 4 strains belonging to type 1 (O157:H7, O103:H2, 0111: H-, and O26:H11) (see Supplementary Table 1). These results suggested that the characteristics of the ter operon revealed in this study were not only those found in E. coli isolated in Japan, but also common to E. coli isolated outside Japan.
Strains carrying the ter-type 4 operon showed significantly lower MICs of potassium tellurite compared to the other three types. The function of Ter proteins has been suggested to involve their mutual association, as they interact at the interface of the inner plasma membrane and the cytosol (Valkovicova et al., 2013;Turkovicova et al., 2016). The nucleotide and amino acid sequences of genes on the ter-type 4 operon were significantly different from those of the other three types ( Supplementary  Figure 1), suggesting that these differences directly affected the functionality of Ter proteins. Most of the ter-type 4 operons were found in isolates from wild deer (94%, 16/17). Of the isolation sources used in this study, deer were the only animals from the wild environment, and their feces were collected from a wide area ranging from Hokkaido in the north, to Kyushu in the south of Japan. In addition to tellurite resistance, the ter operon is also involved in resistance to infection by various bacteriophages and to membrane-pore forming colicins (Whelan et al., 1995(Whelan et al., , 1997Alonso et al., 2000), tolerance to oxidative stresses (Valková et al., 2007), resistance to phagocytosis by macrophages (Ponnusamy and Clinkenbeard, 2015), the ability of cells to adhere to epithelial cells in vitro (Yin et al., 2009), and filamentous bacterial cellular morphology (Whelan et al., 1997). In addition, a ter homolog from Clostridium acetobutylicum showed resistance to methyl methane sulfonate and mitomycin C when expressed in an E. coli recA mutant (Azeddoug and Reysset, 1994). Bacterial acquisition of the ter operon may thus be beneficial for colonization of the host intestinal tract in the presence of various stresses and attacks. Although the reason for the strong relationship between the ter type-4 operon and deer is currently unknown, food and water in the wild environment, and the wild deer gut microbiome formed by their effects, may select strains carrying this type of ter-operon.
The tellurite-resistant phenotypes were variable within each type. For example, five strains of Og177: Hg25 with the tertype 1 operon showed MICs of potassium tellurite ranging from 16 to 128 µg/mL. Furthermore, the amino acid sequences of the ter-type 1 operon genes of major STEC O26:H11 (11,368 strain) was identical to those of Og121:Hg19 (TS23_1 strain) (data not show). However, the resistance level of the latter strain (16-32 µg/mL of potassium tellurite) was roughly lower than that of the former (64-128 µg/mL). A previous study has also shown that even O157:H7 serotype strains differ in their MICs of potassium tellurite (64-1,024 µg/mL) (Taylor et al., 2002). The above results indicate that the ter operon as well as other genetic factors influence tellurite resistance.

CONCLUSION
This study compared the details of ter operons responsible for tellurite resistance among various serotypes and sources of E. coli. Comparative analysis clearly distinguished the ter operon sequences into four subtypes, namely, ter-type 1-4 operons. The amino acid sequences of the ter-type 4 operon genes were distant from those of the other types, further, strains with the ter-type 4 operons showed significantly lower MICs compared to the other types. Furthermore, genomic analysis using long reads indicated the structural details of the integrating elements carrying the ter operons and insertion sites on the chromosome. Genomic comparison and phylogenetic analysis showed that at least three highly conserved elements were spread horizontally within E. coli species. Type 2 was found to be located on plasmids distributed in E. coli as well as in some Enterobacteriaceae species. The tertype was also closely related to the source of isolation, the cause of which was unknown; however, types 2 and 4 were associated with chickens and deer, respectively. This study provided new insights related not only to genetic characteristics of the ter operons in E. coli, but also to phenotypic and ecological differences that may be related to them. In the future, by clarifying the unknown function of the Ter proteins, it is expected that the relationship between the diversity of the ter operon and the phenotypic and ecological characteristics revealed in this study will be clarified.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
TN, AI, and SI conceived and designed the experiments. TN performed the experiments. TN, TK, and TT analyzed the data. TN and AI contributed to reagents, materials, analysis tools, critical revision of the manuscript for important intellectual content, and wrote the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
This research was partially supported by AMED under grant no. JP20fk0108065.