Background

Shigella species are the Gram-negative bacteria that cause shigellosis – a leading cause of bacillary dysentery in developing countries. Annually, 125 million cases of endemic shigellosis occur in Asia alone and children under 5 years of age are at the highest risk of illness and death [1]. Among the four species of Shigella, S. flexneri is the primary cause of shigellosis in developing countries accounting up to 66% of all Shigella species infections [2]. There are 19 different S. flexneri serotypes based on their antigenic determinants present on the O-antigen of the outer membrane lipopolysaccharide (LPS) [3]. The O-antigen modification that results in serotype conversion is mediated by bacteriophages due to addition of glucosyl and/or O-acetyl groups and by plasmid due to addition of phosphoethanolamine groups to one or more sugars of the O-antigen [4]. The O-antigen modification promotes bacterial invasion and evasion of innate immunity; both of which are essential for S. flexneri virulence and this might have contributed to the emergence of serotype diversity [5]. The protective immune response to S. flexneri mainly targets against the O-antigen hence making O-antigen as a major candidate for vaccine development [6]. However, identification of several novel S. flexneri serotypes in the recent years has complicated any potential O-antigen based vaccine development for Shigella [7].

The current approaches to Shigella vaccines consider S. flexneri serotypes 2a and 3a as the major candidates for vaccine development based on animal studies suggesting cross-protection against majority of S. flexneri serotypes [7]. However, S. flexneri serotypes 2a and 3a have not yet been shown to provide cross-protection against S. flexneri serotype 1c strains [8]. S. flexneri serotype 1c, also known as 7a serotype, was first isolated in Bangladesh in the 1980s and was identified as a provisional serotype with unique O-antigen structure comprising of a disaccharide (two glucosyl group) α-D-Glcp-(1➔ 2)-α-D-Glcp linked to O4 of the N-acetyl-glucosamine residue [9]. In the 1990s, an unrelated clone of this serotype was found to be most prevalent S. flexneri serotype accounting for about 40% of S. flexneri isolates in northern Vietnam [10]. Since then, S. flexneri serotype 1c has been isolated and identified in Egypt, Pakistan, China, Canada and the UK [11,12,13]. Therefore, the widespread distribution of S. flexneri serotype 1c highlights the importance of its consideration for Shigella vaccine development strategies.

The glucosylation in O-antigen of S. flexneri is mediated by bacteriophage-encoded genes arranged in an operon known as the gtr cluster which comprises of three genes gtrA, gtrB and gtr(type) [14, 15]. The mechanism of O-antigen modification in serotype 1c is unique compared to other serotypes. The addition of the first glucosyl group is mediated by gtrI cluster found within a cryptic SfI prophage and the addition of second glucosyl group is mediated by separate gtrI cluster designated as gtrIC which came possibly from more distantly related bacteriophage [16]. So far, little is known about S. flexneri serotype 1c in terms of its genome organization, pathogenic islands and the evolution. A better understanding of the pathogenic signatures of any pathogen requires the availability of higher-resolution data such as the whole genome sequence.

A large number of the perfectly repeated insertion sequences in Shigella species impair genome assembly using sequences generated by short-read technologies, e.g. Illumina [17, 18]. This limitation can be overcome with the use of a long-read sequencing technology such as single-molecule real-time (SMRT) sequencing of Pacific Biosciences (PacBio). However, the sequence from SMRT technology has higher error rates compared to Illumina platforms [19]. Recently, a hybrid approach of using both long-read and short-read sequencing technologies has been adopted to yield a high-quality de novo assembly of prokaryotic and eukaryotic genomes [20, 21]. To better understand the genome architecture and dynamics of S. flexneri serotype 1c strains, we have sequenced and assembled S. flexneri serotype 1c strain Y394 using hybrid methods of long-read SMRT and short-read MiSeq Sequencing technologies. We present here the unique features of Y394 genome in comparison with ten other publicly available complete genome sequences of S. flexneri (Additional file 1: Table S1) thus providing further insights into the evolution and pathogenicity of this novel serotype.

Results

Genome assembly

Assembly of the long-read SMRT sequences generated a single bacterial chromosome of 4,585,914 bp with 203X coverage. In common with most other S. flexneri strains, Y394 consisted of a large virulence plasmid of 221,307 bp and an additional small plasmid of 10,873 bp (data not shown). The SMRT sequences were refined using the MiSeq short-read data. The hybrid genome assembly generated the final circular Y394 genome of 4,584,634 bp.

General features

The 4,584,634 bp Y394 genome comprised of 4699 coding sequences with guanine-cytosine (GC) content of 50.9% (2,334,162 of 4,584,634 bp). The number of tRNA sequences were 108, and the number of rRNA sequences were 22. The comparison of genome features show that Y394 has similar nucleotide composition and size (Table 1).

Table 1 General features of the Shigella flexneri genomes
Full size table

Genome rearrangements

To examine the genome rearrangement among S. flexneri genomes, we performed the genome-wide alignment of 10 complete S. flexneri genomes with Y394 as a reference. We identified a high number of homologous genomic regions in all the compared genomes indicating high level of sequence similarity among these strains (Fig. 1). The alignment showed varying degree of genome shuffling and recombination events among different serotypes and strains of S. flexneri.

Fig. 1
figure 1

Genome-wide alignment of Shigella flexneri genomes. The horizontal panel of blocks from top to bottom represents the genome sequences of S. flexneri 1c strain Y394, S. flexneri 5b strain 8401, S. flexneri 2a strain NCTC1, S. flexneri 2a strain 301, S. flexneri 2a strain 2457 T, S. flexneri 2a strain 981, S. flexneri 1a strain 0228, S. flexneri 4c strain 1205, S. flexneri Y strain 2003036, S. flexneri Yv strain Shi06HN006 and S. flexneri Xv strain 2002017, respectively. Each syntenical placement of the homologous regions of the genomes is represented as unique coloured blocks. Blocks above and below the centre line depict the orientation of the genomic region compared to S. flexneri 1c strain Y394. The 12 intact prophage regions are represented by red blocks in Y394 panel. The genomes are added sequentially for comparison based on the phylogenetic distances

Full size image

The GC content and GC skew varied within the Y394 genome (Additional file 2: Figure S1), which could be due to genome rearrangements and horizontal gene acquisitions. The nucleotide composition is relatively uniform over the entire bacterial genome and is conserved within related bacterial species. However, regions with anomalous GC content within a bacterial genome is likely due to recently acquired genes from distantly related organisms [22] and is a common feature of S. flexneri strains [20].

Mobile genetic elements

Similar to other S. flexneri completed genomes (Table 1), the total number of predicted insertion sequence (IS) elements in Y394 genome was 345. The ORFs related to IS elements covered approximately 7% (311,229 bp of 4,584,634 bp) of the total genome with IS3 (47%) and IS1 (33%) being the most common IS elements (Additional file 3: Table S2).

We identified 12 different regions of intact prophage distributed across the genome comprising of 400 CDS with an average GC content of 49.62% (Range: 47.73–52.60) accounting up to 8% (365 kb) of the genome. There were several additional regions with incomplete (5 regions, 45.7 kb) and putative prophage regions (5 regions, 86.9 kb) (Additional file 4: Table S3). However, all of these prophages have undergone massive loss of functional genes resulting in bacteriophage genome fragmentation and ultimately resulting into cryptic prophage. Interestingly, both gtrI and gtrIC gene clusters were found to be present in two different prophage regions which were 2 Mb apart. Both these clusters were lacking phage structural components required for their replication and lysis. Another interesting finding was the identification of functional O-acyltransferase B (oacB) gene 6.8 kb upstream to gtrA gene within the gtrI gene cluster. The functionality of oacB gene was confirmed by slide agglutination assay using specific antiserum against a 3/4-O-acetylated RhaIII epitope prepared as described previously [23]. The analysis of gtrI region revealed that the cryptic phage integrated within the proA-adrA locus of the bacterial chromosome similar to several other S. flexneri serotype-converting phages [24]. The amino acid comparison of oacB gene from Y394 showed 100% sequence coverage and 99% identity (with 3 amino acid difference) to that of Sf101 phage which was identified in different strains of the serotype 1c [25]. We also found that all the three genes, gtrA, gtrB and gtrI of the gtrI operon, were highly conserved with only one point mutation (gtrI gene Ieu 196 Phe) as that of prophage SfI [26]. Both gtrI and gtrIC regions comprised of several transposases and phage hypothetical proteins (Fig. 2a and b).

Fig. 2
figure 2

Linear schematic representation of gtr gene clusters in Y394. The direction of arrows indicates the orientation of open reading frames (ORFs). The color code denotes the ORF types: red, hypothetical protein; plum, transposase and IS elements, light green, tRNA; dark green, integrase; gray/black, annotated protein and blue, gtr genes. a gtrI encoding bacteriophage region with oacB gene in Y394. b gtrIC encoding bacteriophage region in Y394

Full size image

Pathogenicity islands in Y394 genome

Pathogenicity islands (PAIs) are the clusters of mobile genetic elements that encode virulence factors and are present in pathogenic bacterial strains but not present in related non-pathogenic strains [27]. So far, three distinct PAIs have been identified in S. flexneri [28]. The first PAI is the bacteriophage-encoded genes involved in serotype conversion, referred as SHI-O. The SHI-O Island in Y394 comprises 14 kb of gtrI gene cluster and 20 kb of gtrIC gene cluster.

The second PAI or SHI-1, characterized in S. flexneri 2a strains, encodes sigA, pic (formerly known as she) set1A and set1B genes [29]. These genes have been shown to be responsible for fluid accumulation in ligated rabbit ileal loops suggesting their role for the watery diarrhoea in Shigella infection [30, 31]. The SHI-1 is absent in Y394 genome as in several other S. flexneri strains of 1a, 2a, 1b, 3b, 4 and 5b serotypes [20, 32, 33]. (Additional file 5: Table S4).

The third PAI of S. flexneri, SHI-2, is a 23.8 kb region downstream of selC locus and contains genes encoding the synthesis and transport of aerobactin, iron acquisition siderophore system, associated with increased virulence of enteric bacteria [34]. The aerobactin operon consists of first four proteins IucA-D for the siderophore which complexes with iron in the host environment. The protein IutA comprises the bacterial receptor for the iron-siderophore complex [35]. The Y394 genome showed similar composition and organization of genes in the SHI-2 PAI as in other pathogenic 2a strains of S. flexneri (Additional file 6: Figure S2).

The Y394 genome also possessed a 24.8 kb putative pathogenic island, which constitutes ipaH genes that encode effector proteins secreted via the type III secretion system as well as several transposases and hypothetical genes, which might have a role in the virulence [36].

The Y394 was found to be resistant to a number of antimicrobial agents including erythromycin, penicillin, trimethoprim/sulphamethoxazole and tetracycline. The observed phenotype was consistent with antibiotic resistant genes present in the Y394 chromosome and the small plasmid. The chromosomal acrA, acrB and tolC genes codes the tripartite antibiotic efflux pump conferring resistance of Y394 to β-lactams [37, 38] that could also result in resistance to penicillin. The erythromycin resistance in Y394 is mediated by chromosomally encoded multidrug resistant efflux pump (MdtE, MdtF and TolC) [20]. The sulphonamide and tetracycline resistance in Y394 are mediated by plasmid-borne sul2 gene and tetA gene, respectively [39, 40].

Pangenome and phylogenetics

The pangenome of S. flexneri was predicted based on the Y394 and other ten publicly available complete genome sequences of S. flexneri. We identified 3720 chromosomal core genes present in each of the sequenced genomes. The total pangenome of S. flexneri consists of 6237 genes with 2517 accessory genes. The “open” pangenome accumulation curve (Additional file 7: Figure S3) indicated that the pangenome size continues to grow as additional S. flexneri genomes are sequenced. The number of unique genes varied among the compared strains with an average of 84 genes (Range 27–246). Y394 genome had the highest number of unique genes (152 genes) after S. flexneri 5b 8401 genome (246 genes) compared with other S. flexneri genomes (Fig. 3).

Fig. 3
figure 3

Distribution of unique genes found in the S. flexneri genomes. The X-axis is the S. flexneri serotypes, and Y-axis denotes the corresponding number of unique genes. The number of unique hypothetical genes within a genome is shaded black

Full size image

The phylogenetic analysis was performed using 329 chromosomal genes found in all the 23 compared complete genomes including the broader Shigella species and representative strains of Escherichia coli, Klebsiella pneumoniae and Salmonella enterica (Additional file 8: Table S5). A total of 61,547 Single-nucleotide polymorphisms (SNPs) were identified within these core genes. The maximum likelihood tree generated using general time reversible (GTR) model with ascertainment bias correction (ASC) and gamma correction (with n = 4 categories) for site rate variation, identified three major phylogroups (Fig. 4). All the S. flexneri strains grouped together into a single cluster forming a phylogroup with other Shigella species and E. coli. The Y394 genome was evolutionarily more closely related to S. flexneri serotype 5b strain 8401 and S. flexneri serotype 2a strain NCTC1 (Additional file 9: Figure S4). The distantly related species K. pneumoniae and S. enterica formed an outgroup from the Shigella phylogroup as expected.

Fig. 4
figure 4

Evolutionary relationship of Shigella flexneri. The maximum likelihood tree of eleven S. flexneri genomes and other Shigella species, rooted with non-Shigella outgroups. Bootstrap support values of 1000 pseudo-replicates were 100% for all the major clades

Full size image

Discussion

This study represents the first complete genome analysis of S. flexneri serotype 1c which is an emerging S. flexneri serotype in developing countries. Analysis of the Y394 genome revealed that mobile genetic elements including bacteriophages, IS elements, PAIs and plasmids are the driving force for bacterial diversification, adaptation and evolution, and they play an important role in pathogen virulence and spread of drug resistance. The analysis also identified unique and highly modified O-antigen structure comprising of three distinct O-antigen modifying gene clusters (gtrI, gtrIC and oacB) which potentially came from three different bacteriophages. The gtrI gene cluster located at the proA locus mediates the addition of the first glucosyl group and the gtrIC gene cluster mediates the addition of the second glucosyl group resulting into the serotype 1c specific O-antigen modification [16]. The oacB gene found within the gtrI gene cluster is responsible for the 3/4-O-acetylation in the O-antigen [25]. This highly diverse O-antigen modification in serotype 1c perhaps enhances the bacterial ability to escape the host defence mechanism. Further, the phage hypothetical proteins conserved within these clusters might have other roles in virulence. Thus, three different serotype-converting phages were acquired by Y394 and over the period have undergone massive DNA deletions resulting into cryptic phages. Hence, phages are one of the major drivers of adaptively significant genetic variation of S. flexneri 1c and might have roles in pathogen virulence as found in several other bacterial pathogens [41, 42].

In addition to the virulence plasmid, S. flexneri requires chromosomal genes for the full array of virulence phenotypes [43, 44]. Although the SHI-1 PAI was absent, there were several hypothetical genes and transposases present in the upstream region of phe tRNA gene, the integration site of SHI-1 PAI. The ability of the SHI-1 PAI to undergo spontaneous and precise excision via site-specific recombination [45] suggests that Y394 genome might have lost its SHI-1 region during its evolution to incorporate more important genes. The identification of the functions of these acquired genes will be interesting for further studies.

The accumulation of antimicrobial resistance determinants via horizontally acquired plasmids demonstrates how well 1c serotypes are adapted to the evolutionary pressures. The multidrug resistance in S. flexneri serotype 1c clinical isolates is not uncommon [11]. However, the resistance to previously efficacious first-line drugs including sulphonamides, tetracycline and trimethoprim/sulfamethoxazole used in paediatric cases is of concern [46]. The spread of multidrug resistant S. flexneri strain is of greater concern in developing countries due to limited laboratory settings for antibiotic susceptibility testing and unrestricted antibiotic use without proper prescription, resulting into frequent treatment failures and economic burden to underprivileged patients.

Our analysis revealed that S. flexneri is composed of phylogenetically distinct lineages and their genomes are relatively stable compared to other members of enterobacteriaceae family, in line with previous findings by Conner et al. [47]. The different serotypes of S. flexneri are classified based on the O-antigen modification resulted by the genes acquired by horizontally transmissible genetic elements such as bacteriophage and the plasmids. Therefore, it is apparent that the phylogeny based on the core genes cannot segregate the S. flexneri serotypes.

Acquisition of new traits by horizontal gene transfer enables bacterial pathogens to adapt to evolutionary pressures. The cryptic prophages constitute a significant amount of Y394 genome with several hypothetical genes in them. Many prophages carry additional cargo genes in their genomes which are not necessary for the phage but can change the virulence phenotype or fitness of the lysogen leading to the emergence of new pathogens or epidemic clones [41]. As the advancement in sequencing and molecular methods, the list of phage-encoded virulence genes is rapidly growing. Several bacteriophages encoded virulence factors have been identified that play a role in different stages of the bacterial pathogenesis including toxin production, host epithelial cell invasion, adhesion, intracellular survival and altering of antigenicity [42, 48, 49]. The clear understanding of the role of these unknown bacteriophage genes in the survival of S. flexneri serotype 1c in the human host can pave the way to the identification of potential attenuation targets and vaccine candidate antigens.

Conclusions

Although a dozen of S. flexneri strains have already been sequenced, for the first time, we have sequenced the complete genome of S. flexneri serotype 1c (strain Y394) using hybrid sequencing approach and produced a high-quality reference genome. S. flexneri serotype 1c is an important serotype responsible for more recent shigellosis outbreaks in the developing nations. Our findings showed that the genome of Y394 is highly complex with a large number of unique genes acquired via horizontal gene transfer, which might be important for the pathogenesis and virulence of the pathogen. The overall genomic organization, gene contents and order are similar in all the S. flexneri genomes. However, the collinearity is disrupted by transposons mediated inversions. There are variations in the PAIs, more notably being SHI-O (containing O-antigen) in different S. flexneri genomes. The genes present in both the chromosome and the plasmid confer the resistance of the Y394 to antimicrobial agents. The identification of several hypothetical genes in the PAIs and putative bacteriophage regions warrants future investigation for determining the role of these unknown genes in relevance to pathogen’s virulence and survival in the host environment.

Methods

Bacterial strain and laboratory methods

S. flexneri serotype 1c strain Y394 was a clinical isolate from Bangladesh and kindly gifted by Nils I. A. Carlin [9]. The Y394 was grown aerobically (180 rpm) at 37 °C in Luria Bertani broth (LB). Bacterial DNA was extracted using the Genomic Tip 100/G (Qiagen) according to the manufacturer’s instructions.

The antibiotic susceptibility pattern of Y394 was determined using disk diffusion method (Kirby-Bauer) [50]. The antibiotic discs (Oxoid, UK) used in this study were ampicillin (10 μg), cefoxitin (30 μg), chloramphenicol (30 μg), erythromycin (30 μg), kanamycin (30 μg), nitrofurantoin (300 μg), penicillin (1 U), tetracycline (30 μg) and trimethoprim/sulfamethoxazole (1.25/23.75 μg).

Whole genome sequencing

Genome sequencing of Y394 was performed using both SMRT sequencing (PacBio) and MiSeq sequencing (Illumina). The SMRTbell Template Prep Kit 1.0 (PacBio) was used for SMRT sequencing library preparation of about 20 kb insert size. The sequencing was performed using PacBio RSII sequencing system. The Nextera XT DNA library preparation kit (Illumina) was used for MiSeq v3 300 bp paired-end sequencing. Both the sequencing and library preparation was performed at the Ramaciotti Centre for Genomics, The University of New South Wales, Australia.

Genome data and assembly

The PacBio sequencing was performed using 240 min movies (Using P6-C4 chemistry) that generated 77,493 individual reads (sub reads) of DNA fragment with N50 of 20.6 kb. A de novo assembly of these reads was performed with HGAP.3 (Pacific Biosciences) on the SMRT Analysis Pipeline version 2.3.0 [51]. The Y394 genome was obtained as a contiguous sequence of 4,585,914 bp with 203X coverage.

The MiSeq sequencing generated 2 × 338,140 paired-end reads of Y394 genome with the average fragment size of 291 bp (range 55–301). The quality of the reads was assessed using FastQC v0.11.5 [52]. The initial 20 bp and beyond 200 bp of each read were trimmed to obtain high quality bases using Trimmomatic v0.36 [53]. The assembly of the trimmed reads was performed using the VelvetOptimiser Script [54].

BWA-MEM [55] was used to map the trimmed MiSeq reads against the PacBio based genome assembly followed by the use of Pilon for PacBio sequence assembly improvement [56, 57]. The sequence assembly pipeline adapted from [58] is outlined in Additional file 10: Figure S5.

Sequence analysis

The annotations of the final Y394 genome as well as other completed genomes of S. flexneri included in this study were performed using Rapid Annotation of Prokaryotic Genomes (prokka) [59]. The prokka annotation was re-annotated with PHASTER (PHAge Search Tool - Enhanced Release) database [60] to find the bacteriophage-encoded regions in the Y394 chromosome. The genome-scale alignments were performed using Mauve alignment Tool [61] and CG View Server [62]. The IS elements were identified using ISsaga [63]. The antibiotic resistance profile was generated using ARDB-Antibiotic Resistance Genes Database [64].

The pangenome of S. flexneri was predicted based on the completed genomes of different strains of S. flexneri. All these genomes were re-annotated using prokka to avoid biases in the comparisons due to different annotations. The pangenome analysis was performed using Roary Pipeline [65]. The phylogenetic tree was constructed based on the 329 core genes present in all the compared S. flexneri genomes and non-S. flexneri genomes including representative strains of other Shigella species, E. coli, K. pneumoniae and S. enterica. The 61,547 SNP sites were identified and concatenated to form a multiple alignment of SNPs for phylogenetic analysis using the package SNP-sites [66]. The best-fit substitution model was determined (GTR + ASC + G4) and used for constructing the maximum likelihood tree using the package IQ-tree [67]. The bootstrap analysis was performed using 1000 randomizations. The image files were generated using Snap Gene Viewer (Version 3.3.1), GraphPad Prism (Version 5.01) and Fig Tree (version 1.4.3) [68].