Characterization of integrative and conjugative elements carrying erm(B) and tet(O) resistance determinants in streptococcus uberis isolates from bovine milk in Chiba prefecture, Japan: CompArative GEne cluster analysis toolbox with ICEfinder

Abstract

Objective

We aimed to characterize integrative and conjugative elements (ICEs) in antimicrobial resistant Streptococcs uberis isolates from bovine milk in Chiba, Japan, based on whole-genome sequence (WGS) data.

Results

Of the 101 isolates, we found the 36 isolates harboring erm(B)–tet(O), showing resistance to macrolides–lincosamides–tetracyclines. The 22 isolates were randomly selected and subject to WGS determination. The genomes measured 1.991–2.517 Mbp, with G + C contents of 35.8–36.9%. We used ResFinder–ICEfinder (web-based applications) to search for the antimicrobial resistant genes and ICEs. ResFinder detected combined erm(B)–tet(O)–ant(6)-Ia at the identical contig in each WGS. ICEfinder detected ICEs belonging to the same contigs, which contained erm(B)–tet(O)–ant(6)-Ia complete or partial sequences. Detection of putative ICEs using comparative genomic analysis was performed with identification of other streptococcal ICE resembling S. uberis ICEs. Using comparative genomic analysis (a reference WGS in NZ01 strain), putative ICE base size in UB37 isolate was 77,386-bp that was identical in other 13 isolates. Another similar streptococcal ICE was S. suis ICEnsui78–tet(O)–erm(B) mobile element. For ICE characterization in S. uberis with WGSs, a comparative genomic analysis is required with use of ICEfinder and other annotation tools.

Introduction

Mobile genetic elements (MGEs) play a major role in horizontal gene transfer among microorganisms. Integrative and conjugative elements (ICEs) of MGEs are self-transmissible in bacteria and are characterized by both integrative and conjugative features [1]. ICEs are mosaic elements with both bacteriophage- and plasmid-like characteristics that integrate into and replicate with host cell chromosomes [2]. Additionally, ICEs transfer their antimicrobial resistance (AMR) determinants as well as the genes involved in their mobility/regulation/maintenance. Many ICEs are demonstrated among the genus Streptococcus [3]. ICEs belonging to Tn916-Tn1545 family transfer the widely disseminated tetracycline-resistance determinant tet(M), along with additional determinant erm(B) [4]. erm(B) gene encodes methylase that generates a posttranscriptional methylation of 23 S rRNA, leading to cross-resistance to macrolides, lincosamides, and streptogramin B antimicrobials [5].

Bovine clinical/subclinical mastitis in dairy cattle affect udder health, milk quality, and milk production, resulting in economic losses in dairy farms. Major causative microorganisms of the mastitis contain Escherichia coli, Staphylococcus aureus [6], coagulase-negative staphylococci, and contagious/environmental streptococci [7]. Contagious Streptococcus is S. agalactiae, whereas environmental streptococci contain S. dysgalactiae, S. canis, and S. uberis [8, 9].

S. uberis is isolated from environmental sources (soil, pasture, bedding materials, and bovine feces) and is present on the skin surface of dairy cows [8]. This microorganism develops bovine (sub)clinical mastitis during lactating/non-lactating periods after direct contact with the teat apex [10]. Molecular epidemiological investigation regarding S. uberis-associated clinical mastitis in dairy herds described that either predominant or a limited number of isolates may produce intramammary infections or transmission among cows (including potential transmission of the isolates via milking machine/environment) [11].

Macrolides, lincosamides, and beta-lactams are administered for treatment of S. uberis-associated mastitis [12]. Excessive usage of antimicrobials in dairy herds can result in elevated AMR among mastitis bacteria [13]. Zhang et al. [12] reported AMR phenotypes/genotypes among S. uberis isolates from bovine mastitis in Thailand. The isolates were resistant to tetracycline (82.0%), ceftiofur (19.3%), and erythromycin (8.3%). Prevalent AMR genes were tet(M) (87.3%), erm(B) (66.2%), and blaZ (6.6%). We have showed AMR patterns of S. uberis from bovine milk in Japan [14]. AMR monitoring of bovine mastitis-associated S. uberis may support an antimicrobial stewardship program for dairy farms.

Based on the findings by polymerase chain reaction (PCR) method, Haenni et al. [15] documented diverse and mobile features of ICEs among bovine S. uberis isolates. We searched for related articles by entering the keywords “streptococcus uberis, integrative and conjugative element, whole-genome” in PubMed database (https://pubmed.ncbi.nlm.nih.gov/) [16]. There is limited literature available on this topic, although sequence characterization and novel insights into bovine mastitis-associated S. uberis in dairy herds have been reported [17]. The aim of our work is to characterize ICE characterization in S. uberis AMR isolates from bovine milk based on whole-genome sequence (WGS) data.

Methods

AMR phenotyping/genotyping

All bovine milk-origin isolates (n = 106) identified based on mass spectrometry results were provided by Sanritsu Zelkova Veterinary Laboratory during Mar–Oct 2022. This laboratory determined minimum inhibitory concentrations of 14 antimicrobials (penicillin-G/ampicillin/minocycline/erythromycin/azithromycin/clindamycin/levofloxacin/chloramphenicol/cefotaxime/ceftriaxone/cefepime/cefozopran/meropenem/vancomycin) using broth microdilution method (MICroFAST Panel Types 7 J for Streptococcus spp.; Beckman Coulter Inc., Tokyo, Japan) recommended in Clinical and Laboratory Standards Institute (CLSI) M100-S26 document for alpha-hemolytic streptococci [18]. Quality control was performed using Enterococcus faecalis American Type Culture Collection (ATCC) 29,212, Streptococcus pneumoniae ATCC 49,619, and S. uberis ATCC 700,407 strains with the antimicrobial susceptibilities. To determine susceptibility/resistance to minocycline, we used tetracycline breakpoints in accordance with the CLSI guidelines.

One isolate per host was stored at − 70 °C to − 80 °C in Kitasato laboratory. We performed PCR methods to differentiate S. uberis and S. parauberis according to PCR fragments of 16 S and 23 S rRNA genes [19]. One hundred one isolates were analyzed. AMR genotyping [blaZ–erm(A)–erm(B)–mef(A)–lnu(B)–lnu(D)–tet(M)–tet(O)–tet(K)–tet(L)–tet(S)] was conducted using PCR assay [20, 21].

Randomized selection of S. uberis isolates with erm(B)–tet(O)

We found the isolates harboring both erm(B) and tet(O) (n = 36): all isolates indicated resistance to macrolides, lincosamides, and tetracyclines (the AMR rates of 33.7%, 54.5%, and 48.5%) [14]. Combined erm(B)–tet(O) genotype seems to be a representative marker to find ICEs in corresponding isolates, because both genes are located within MGEs [15]. The twenty-two isolates with erm(B)–tet(O) were selected based on random sampling numbers using Excel application. S. uberis ATCC 700,407 having WGS (https://genomes.atcc.org/genomes/26250c65fb774abc) at ATCC site was used as a reference strain. The enrolled strains’ list is shown in Supplementary Table 1.

WGS determination

DNAs from the twenty-two isolates were extracted as previously reported [22]. Whole-genome sequencing was carried out on a DNBSEQ-G400RS platform (MGI-Tech, Tokyo, Japan) using DNA Nanoball technology [23]. Sequencing library was generated using MGIEasy FS DNA Library Prep Set. Paired-end runs were performed with a read length of 2 × 150-bp.

We assessed read lengths, ambiguous bases, and quality scores for quality control. The reads were trimmed using quality trimming tool in CLC Genomics Workbench (v.8.0.2) with default parameters (quality limit and maximum number of ambiguous bases). De novo assembly was conducted using CLC Genomics Workbench with modified parameters, wherein the minimum contig length was set to 800-bp. Draft genome sequences were annotated using DDBJ Fast Annotation and Submission Tool (DFAST; https://dfast.nig.ac.jp) [24]. We determined assembly metrics and annotated features including genome size, number of contigs, average coverage, N₅₀, numbers of coding DNA sequences (CDSs)/tRNAs/rRNAs/clustered regularly interspaced short palindromic repeats (CRISPR), G + C content, and coding ratio.

Additionally, the sequence type/clonal complex were determined using S. uberis PubMLST Isolate database (https://pubmlst.org/bigsdb?db=pubmlst_suberis_isolates). We also constructed the phylogenetic tree of 22 isolates and ATCC 700,407 as previously reported [25].

Web-based applications

We used ResFinder v.4.1 (https://cge.food.dtu.dk/services/ResFinder/) and ICEfinder v.1.0 (https://bioinfo-mml.sjtu.edu.cn/ICEfinder/index.php) with WGSs, to comprehensively search for related AMR genes and ICEs [25, 26]. Among the ICEs found, we selected the ICE containing erm(B)–tet(O) genes as a candidate from each isolate for further characterization.

Detection of putative ICEs using comparative genomic analysis

Flanking products of each candidate ICE were confirmed based on DFAST annotations. We selected a complete S. uberis WGS sequence without combined erm(B)–tet(O) as a reference. This WGS needed to possess flanking products identical to those of selected ICE based on the ICEfinder results. When we subtracted flanking products in reference WGS from corresponding products in candidate WGSs (comparative genomic analysis), we could detect putative ICEs containing erm(B)–tet(O). The subtraction procedure was based on concept “integration” resulting from ICE definition.

The putative ICE base size and product organization were determined for each WGS. To conduct homologous gene cluster searches and to generate publication-quality figures of gene cluster comparison between reference and candidate WGSs, we applied CompArative GEne Cluster Analysis Toolbox (CAGECAT, https://cagecat.bioinformatics.nl) consisting of two components of cblaster (for searches) and clinker (for figures) [27,28,29]. A Prokaryotic Genomes Annotation Pipeline (PGAP) in National Center for Biotechnology Information (NCBI) was applied for re-annotation. To specify ICE components, ICEscreen software was also applied [30]. To speculate protein function similarities, we used basic local alignment search tool x (blastx, translated nucleotide to protein homology search) on NCBI.

We found out where ICEUB37 was inserted into the genome selected.

Searching for other streptococcal ICE resembling S. uberis ICEs

Product organization of each determined ICE was inserted into homology search box on NCBI site. Other streptococcal ICEs with high-percent identity with the inserted query were extracted. One ICE showing the high-percent identity with inserted ICEs in our 22 isolates was specified. Using clinker, we constructed a visualized and interactive figure among the specified ICE and our ICEs.

Results

Assembly metrics and annotated features

WGS assembly metrics and annotated features (genome size, number of contigs, average coverage, N₅₀, numbers of CDSs/tRNAs/rRNAs/CRISPR, G + C content, and coding ratio) are shown in Supplementary Table 2.

Additionally, the sequence type/clonal complex are shown in Supplementary Table 2. Of 22 strains, twenty belonged to CC996, suggesting that this population is closely related. The constructed phylogenetic tree of 22 isolates and ATCC 700,407 is shown in Supplementary Fig. 1.

Web-based application results

Web-based application results with WGSs (AMR genes and ICEs) are shown in Supplementary Table 3. There were combined genes erm(B)–tet(O)–ant(6)-Ia located at identical contig for each WGS. We observed other AMR genes [lnu(C), lsa(E)–lnu(B), or lnu(C)–fosB] at different contigs of WGSs in three isolates (UB25, UB78, or UB97). ICEfinder detected ICEs belonging to the same contigs containing combined erm(B)–tet(O)–ant(6)-Ia complete or partial sequences. All ICEs included erm(B)–ant(6)-Ia complete sequences, whereas some possessed tet(O) complete sequence: others had tet(O) partial sequence (Supplementary Table 3). Sometimes ICEfinder does not delimit the ICEs correctly and misses some part of the ICE.

Detection of putative ICEs using comparative genomic analysis

Detection approach of putative ICEs using comparative genomic analysis is shown in Fig. 1. As a reference WGS, we selected a complete S. uberis sequence (nucleotide accession no. CP022435.1) in bovine mastitis-associated NZ01 strain from New Zealand in 2014. UB37 isolate was chosen as an analyzed WGS. NZ01 and UB37 WGSs shared identical flanking products of 8-oxo-dGTP diphosphatase and RlmD [23 S rRNA (uracil(1939)-C(5))-methyltransferase].

Fig. 1

Detection approach of putative Integrative and conjugative elements (ICEs) using comparative genomic analysis. As a reference, we applied a complete genome sequence (nucleotide accession no. CP022435.1) in bovine NZ01 strain from New Zealand in 2014: UB37 isolate was chosen. Similar/identical product comparisons between reference and candidate WGSs are shown in the same colors. Black gradations in interspace among the corresponding genes indicate percent identity. NZ01 and UB37 WGSs possessed identical flanking products of 8-oxo-dGTP diphosphatase and RlmD [23 S rRNA (uracil(1939)-C(5))-methyltransferase]. Putative ICE base size was 77,386-bp. Product organization in ICE was determined. The core products are shown in red letters; the antimicrobial resistance products are shown in blue letters. We performed the re-annotation for ICEUB37 using a Prokaryotic Genomes Annotation Pipeline. To specify ICE components, ICEscreen was applied. To speculate protein function similarities, we used basic local alignment search tool x; the protein functions are shown in brackets

Using comparative genomic analysis, the putative ICE base size of UB37 WGS was considered 77,386-bp, and the product organization of ICEUB37 was determined (Fig. 1). This ICE contained site-specific DNA-methyltransferase, serine integrase, relaxase MOBP (PF03432), MobC family plasmid mobilization relaxosome protein, SAG1252 family conjugative relaxosome accessory protein, SNF2-related protein, phage tail tip lysozyme, conjugal transfer protein TrbL, DNA (cytosine-5-)-methyltransferase, and replication initiator protein A. Using ICEscreen, we confirmed that ICEUB37 included a component set consisting of VirB4 and coupling protein VirD4. Its virB4 and virD4 sequences (sizes 2,352-bp and 1,818-bp) were similar with those of S. suis D9 complete genome (nucleotide accession no. CP002641.1) belonging to the Tn5252 (Supplementary Fig. 3). Furthermore, using blastx, we found that conjugal transfer protein TrbL has VirB6 function and that phage tail tip lysozyme has in part peptidoglycan hydrolase PcsB in Streptococcus.

Supplementary Fig. 2 indicates where ICEUB37 was inserted into the NZ01 genome.

Similarities/differences in ICE results (base start to end and size) between comparative genomic analysis and ICEfinder are shown in Table 1. There were two prevalent ICE base sizes by comparative genomic analysis consisting of 77,386-bp (UB7/UB15/UB25/UB36/UB37/UB41/UB56/UB58/UB63/UB77/UB81/UB82/UB89/UB97) and 76,735-bp (UB68/UB85/UB92). In contrast, we found three prevalent ICE base sizes by ICEfinder containing 48,105-bp (UB15/UB25/UB37/UB38/UB56/UB63/UB82/UB89/UB97), 49,736-bp (UB7/UB36/UB41/UB58/UB81), and 85,330-bp (UB68/UB85).

Table 1 Similarities/differences in ICE results between ICEfinder and comparative genomic analysis

However, we obtained partial ICE base size (> 49,075-bp) in UB99. The accurate estimation was impossible because of the analyzed position at contig terminal (Table 1/Fig. 2) (see the Limitations).

Fig. 2

Putative identification of other streptococcal Integrative and conjugative element (ICE) resembling S. uberis ICEs (UB37/UB68/UB23/UB99), with S. uberis NZ01 WGS. Similar/identical product comparisons between reference and candidate WGSs are shown in the same colors. Black gradations in interspace among the corresponding genes indicate percent identity. Another similar streptococcal ICE was Streptococcus suis strain STC78 ICEnsui78–tet(O)–erm(B) mobile element (nucleotide accession no. ON944185.1), based on the similar product arrangements between UB37/UB68/UB23/UB99 ICEs and ICEnsui78. The core products are shown in red letters; the antimicrobial resistance products are shown in blue letters. We performed the re-annotation for ICEUB37 using a Prokaryotic Genomes Annotation Pipeline. To specify ICE components, ICEscreen was applied. To speculate protein function similarities, we used basic local alignment search tool x; the protein functions are shown in brackets

Putative identification of other streptococcal ICE resembling S. uberis ICEs

Putative identification of other streptococcal ICE resembling S. uberis ICEs (UB37/UB68/UB23/UB99) is shown in Fig. 2. Another similar streptococcal ICE was Streptococcus suis strain STC78 ICEnsui78–tet(O)–erm(B) mobile element (nucleotide accession no. ON944185.1; 55,758-bp), based on the similar product organization between UB37/UB68/UB23/UB99 ICEs and ICEnsui78. In contrast, this element did not possess arrangement of Ant(6)-Ia.

Discussion

Firstly, ICEfinder was applied to comprehensively search for ICEs. Some ICEs possessed tet(O) complete sequence, whereas others had tet(O) partial sequence (Supplementary Table 3), although phenotypic resistance to minocycline was confirmed in all isolates (Supplementary Table 1). This suggests that ICEfinder did not consider the entire ICE. Secondary, we performed putative ICE detection using comparative genomic analysis (a reference WGS in NZ01 strain). Twenty-one ICEs were determined (Table 1/Fig. 2). We could not obtain the accurate results of UB99 ICE because of the analyzed position at contig terminal. Therefore, we should construct complete WGSs using short- and long-read sequencing as we previously reported complete WGSs of Streptococcus canis strains from dogs [31].

The facts that ICEUB37 can be found in other streptococci is significant topics for discussion. Huang et al. [2] reported genomic organization of ICESa2603 (nucleotide accession no. AE009948.1; 54,349-bp) in S. agalactiae strain 2603 V/R. ICESa2603 comprised three proteins of SAG1250 (similar with prokaryotic DNA relaxase), SAG1251 (similar with MobC of prokaryotic plasmid), and SAG1252: ICEUB37 contained SAG1250 family conjugative relaxase [relaxase MOBP (PF03432)], MobC family plasmid mobilization relaxosome protein, and SAG1252 family conjugative relaxosome accessory protein. We found the similarity in ICE component involved with mobilization between ICESa2603 and ICEUB37. A structure, which was very similar with ICEUB37 [66,036/66,071 (99.95%)], was also found in S. ruminantium WGS (nucleotide accession no. AP025333.1; 67,641-bp spanning nucleotides from 1,168,946 to 1,236,586), using homology search in NCBI nucleotide database. S. ruminantium is the causative agent of several bovine and ovine diseases and S. ruminantium-related sheep mastitis outbreak has recently been reported from Italy [32]. Another structure similar with ICEUB37 [33,405/34,281 (97.44%)] was observed in S. pluranimalium WGS (nucleotide accession no. CP120510.1; 34,273-bp spanning nucleotides from 26,385 to 60,657). The other structure similar with ICEUB37 [27,974/28,781 (97.20%)] was also confirmed in S. macedonicus WGS (nucleotide accession no. CP113440.1; 28,775-bp spanning nucleotides from 1,544,153 to 1,572,927).

Another streptococcal ICE resembling S. uberis ICEs was S. suis ICEnsui78–tet(O)–erm(B) [33]. However, S. suis ICEnsui78 seems incomplete, because of lacking VirB4, coupling protein VirD4, and replication initiator protein A in the ICE. This could be possibly resulting from shotgun WGS using Illumina that did not fully resolve the ICE. Additionally, we analyzed the conjugation family [34] to which ICEUB37 belongs. This ICE fits into the Tn5252 (conjugation family) (Supplementary Fig. 3).

In conclusion, for full ICE characterization in S. uberis with WGSs, a comparative genomic analysis is required with use of ICEfinder and other annotation tools.

Limitations

Characterizing ICEs of selected S. uberis strains (n = 22) only by short-read WGS technology may be a limitation as repetitive sequences may be dismissed. Given the significant clinical implications of this research, it is recommended that whole-genome sequencing needs to be conducted using long-reads to enable more meaningful linear comparisons. We may consider requesting an extension for further modifications to solve the limitations of this study. Creations of hybrid genomes on two platforms making long- and short-read WGSs should be performed to completely identify ICEs. To the best of our knowledge, conjugation experiments using S. uberis have not been reported. We need to perform in vitro conjugation experiments to confirm ICE transfer between selected donor and recipient strains. Detailed information (antimicrobial dosing and duration) should be clarified for the relationships between ICEs and their clinical implications.