Comparative Genomics of Escherichia coli Isolated from Skin and Soft Tissue and Other Extraintestinal Infections

ABSTRACT Escherichia coli, an intestinal Gram-negative bacterium, has been shown to be associated with a variety of diseases in addition to intestinal infections, such as urinary tract infections (UTIs), meningitis in neonates, septicemia, skin and soft tissue infections (SSTIs), and colisepticemia. Thus, for nonintestinal infections, it is categorized as extraintestinal pathogenic E. coli (ExPEC). It is also an opportunistic pathogen, causing cross infections, notably as an agent of zoonotic diseases. However, comparative genomic data providing functional and genetic coordinates for ExPEC strains associated with these different types of infections have not proven conclusive. In the study reported here, ExPEC E. coli isolated from SSTIs was characterized, including virulence and drug resistance profiles, and compared with isolates from patients suffering either pyelonephritis or septicemia. Results revealed that the majority of the isolates belonged to two pathogenic phylogroups, B2 and D. Approximately 67% of the isolates were multidrug resistant (MDR), with 85% producing extended-spectrum beta-lactamase (ESBL) and 6% producing metallo-beta-lactamase (MBL). The blaCTX-M-15 genotype was observed in at least 70% of the E. coli isolates in each category, conferring resistance to an extended range of beta-lactam antibiotics. Whole-genome sequencing and comparative genomics of the ExPEC isolates revealed that two of the four isolates from SSTIs, NA633 and NA643, belong to pandemic sequence type ST131, whereas functional characteristics of three of the ExPEC pathotypes revealed that they had equal capabilities to form biofilm and were resistant to human serum. Overall, the isolates from a variety of ExPEC infections demonstrated similar resistomes and virulomes and did not display any disease-specific functional or genetic coordinates.

ease conditions were detected. This study provides functional molecular infection epidemiology insight into SSTI-associated E. coli compared with ExPEC pathotypes. KEYWORDS Escherichia coli, genomics, sepsis E scherichia coli is one of the most important agents of extraintestinal infections, with the potential to cause infection in almost any anatomical site. It can be grouped into pathotypes, such as uropathogenic E. coli (UPEC), septicemia-associated E. coli (SePEC), skin and soft tissue infection (SSTI)-associated E. coli, neonatal-meningitis-causing E. coli (NMEC), and avian-pathogenic E. coli (APEC) (1,2). Urinary tract infections (UTIs) alone account for about 25% of all cases of septic shock, while soft tissue infections represent 10% of reported severe sepsis cases (3,4). Clinical stages of UTIs vary from nonsymptomatic acute infection to severe septic shock. In the latter case, the infection generally progresses from the urinary tract to the bladder (cystitis) and from the bladder to the kidneys (pyelonephritis) and finally into the bloodstream (urosepsis). Skin and soft tissue infections (SSTIs) are major bacterial infections, often self-limiting but in severe cases requiring hospitalization and parenteral antibiotic therapy. With the indiscriminate use of antibiotics and emergence of multidrug-resistant (MDR) organisms, such as sequence type 131 (ST131) and carbapenem-resistant strains, treatment has become quite challenging (5)(6)(7). Despite the fact that SSTI E. coli is an important pathotype, to our knowledge only UTI E. coli isolates isolated in India have been characterized (8,9). This study provides a functional molecular infection insight for E. coli associated with SSTI. Phylogenetic relationships, virulence profiles, antibiotic resistance patterns, and functional attributes were determined, and whole genomes were sequenced and compared with septicemia and pyelonephritis isolates from the same setting of endemicity in India.

RESULTS
Phylogenetic groups and antimicrobial resistance. The majority of the isolates included in this study were assigned to virulence B2 (36%) and D (29%) phylogroups, with 23% in the A group and 12% in the B1 group (Table 1). Of the three disease types, more than half of the isolates fell into B2 and D phylogroups. In total, 73 isolates (93%) were resistant to at least one of six antimicrobial agents, with highest resistance being shown to ciprofloxacin (86%), followed by tetracycline (77%), co-trimoxazole (72%), gentamicin (31%), chloramphenicol (23%), and fosfomycin (5%). Details of all bacterial isolates, including their antimicrobial resistance patterns, are provided in Table S1 in the supplemental material. Overall, 67% of the total E. coli isolates were found to be multidrug resistant (MDR), in which pyelonephritis, septicemia, and SSTI contributed fractions of 0.40, 0.32, and 0.26, respectively ( Table 2). The extended-spectrum-betalactamase (ESBL)-producing E. coli strains were found to comprise as much as 85% of the isolates and were similar in all three disease types. Only 6% of the isolates were metallo-beta-lactamase (MBL) producers. Resistance gene profiling was compatible with the phenotypic observations, with bla CTX-M-15 being the predominant genotype and NDM-1 being a common occurrence among all MBL producers (Tables 1 and 2). The gene sul1 was detected in approximately 71% of the isolates in all three categories, conferring resistance to sulfonamides, while the plasmid-based fluoroquinolone resistance gene [aac(6=)-lb-cr] was detected in 30% and 71% of the pyelonephritis and SSTI isolates, respectively.
Virulence genotypes. Analysis of the virulence genotype fimH showed that it was present in ϳ90% of the isolates in each category, with the exception of SSTI isolates (62%). Interestingly, papC was found to be prevalent in septicemia and pyelonephritis isolates. However, afaB/C and sfaD/E were present in various proportions in the three groups. Of the three toxin types, cvaC was detected, to a lesser extent, in pyelonephritis and septicemia isolates. The prevalence of protectin genes traT and ibeA was the same as that of iron acquisition genes iroN and iucD in all three groups (Table 1).
Whole-genome analysis of resistome and virulome. Twelve strains, comprising four from each of the three disease types, were subjected to comparative genomic analysis to obtain a broader perspective of the complexity of the E. coli disease types. Genome sequence data for four strains were also included from previous and ongoing studies, including reference strains Escherichia coli NA097 and NA114 (10,11), and eight isolates of this study were subjected to whole-genome sequencing (WGS). Wholegenome-based resistome profiles confirmed the strains to be multidrug resistant. All strains demonstrated relatively similar combinations of resistance genes encoding  antibiotic efflux pumps, regulators, antibiotic inactivation, and target modification (Fig. 1b). The virulomes also showed very little pathotype specificity (Fig. 1a). Based on in silico multilocus sequence typing (MLST) results, two SSTI strains (NA633 and NA643) were assigned to the pandemic ST131 clone, while the rest of the strains were assigned to different sequence types (STs), namely, ST38, ST68, ST405, and ST 617 ( Table 3). Results of comparative genomic analyses using the BLAST Ring Image Generator (BRIG) platform showed sequence-type-specific similarities, i.e., all five ST131 strains  had similar arrangements of genomic regions, whereas isolates belonging to other STs had dissimilar regions (Fig. 2a). Clearly, disease-specific genomic features were not detected.
Functional virulence determination. A total of 30 E. coli isolates, comprising 10 from each category, were analyzed for functional virulence determinants. Results revealed biofilm formation capabilities for all isolates when tested using M63 medium and incubation for 48 h at 28°C. Most were strong biofilm formers (Fig. 2c) and also able to grow in human serum (50%). Resistance to serum, i.e., demonstrating no deleterious effect, was variable among the isolates, but the difference was insignificant (Fig. 2b).

DISCUSSION
Infections caused by extraintestinal pathogenic E. coli (ExPEC) are common worldwide. In the present study, we observed that E. coli strains isolated from SSTI, septicemia, and pyelonephritis were predominately B2 and D phylogroups (Table 1), which are considered virulent E. coli (12,13). The observations of this study are similar to those of Petkovsek et al. (14), who reported that SSTI E. coli possessed virulence factors (VFs). Ananias and Yano (15) also showed VFs to be common among different sepsisassociated E. coli (SePEC) strains notably able to invade Vero cell lines.
Recent studies reported a global increase in multidrug resistance (MDR) and extended-spectrum beta-lactamase (ESBL) genotypes among ExPEC strains associated with infections (16,17). In the present study, the majority (67%) of ExPEC isolates were MDR, with CTX-M-15 being the predominant ESBL genotype (Tables 1 and 2). Furthermore, the combination of ESBL and fluoroquinolone resistance was the most frequent, followed by that of CTX-M-15 positivity (CTX-M-15 ϩ ), fluoroquinolone resistance, and co-trimoxazole resistance and that of CTX-M-15 ϩ , fluoroquinolone resistance, cotrimoxazole resistance, and tetracycline resistance ( Table 4). The resistome (Fig. 1b), based on WGS, showed the presence of multiple genes, whereas the MBL phenotype in all isolates was associated with the presence of the NDM gene, reflecting dissemination of NDM carbapenem resistance in India (7).
ExPEC is known to be evolving and disseminating globally via clonal expansion, with clones of similar genetic architectures but also demonstrating strain-specific features (11,18). Here, we observed similarities in virulence and resistance coordinates of strains belonging to sequence types (STs) ST131, ST38, ST68, ST405, and ST617. Interestingly, ST-specific commonality but not disease-specific similarity was observed (Fig. 2a). These results support the conclusion that strains of diverse STs are capable of causing similar infections. We also report for the first time the occurrence in India of E. coli isolates from SSTIs that carry ST131.
In summary, the results of this study indicate that ExPEC isolates associated with pyelonephritis, septicemia, and SSTIs comprise overlapping phylogroups and patterns of virulence, drug resistance, genomic, and functional properties but are not specifically associated with pathotypes.

MATERIALS AND METHODS
Bacterial strains. A total of 78 isolates, including 21 from patients suffering infections associated with skin and soft tissues, were obtained from the D. Y. Patil Hospital, Pune, India, during 2015. Also, 57 E. coli strains isolated during January 2009 to December 2012 were included in the study. All isolates were identified and preserved employing standard laboratory methods of the hospital as previously published (8) and the protocols were approved by Institutional Biosafety Committee (IBSC) of University of Hyderabad, Hyderabad, India.
Preparation of DNA template and E. coli phylogenetic grouping. Template DNA was prepared using the boiling lysis method. In brief, 100 l of the bacterial cultures was boiled at 95°C for 20 min and centrifuged at 6,000 rpm for 10 min. The supernatant obtained was used as the template. All E. coli isolates were classified into four major phylogenetic groups by multiplex PCR using three molecular markers: chuA, yjaA, and tspE4 (19). After electrophoresis (1.5% agarose), gel images were captured and isolates were assigned to phylogenetic groups based on the dichotomous decision tree of the work of Clermont et al. (19). Comparative Genomics of E. coli Isolates ® ESBL screening and antimicrobial resistance. ESBL production was confirmed phenotypically, using guideline M31-A3 of the Clinical and Laboratory Standards Institute (CLSI) (20). Resistance to carbapenems was determined by using Etest (HiMedia), and susceptibility was defined by breakpoints per CLSI (20). All E. coli isolates were tested for resistance to six major classes of non-␤-lactam antibiotics, employing standard disc agar diffusion on Muller-Hinton agar (HiMedia) (20). Six antimicrobial discs (HiMedia) were used: ciprofloxacin (30 g), chloramphenicol (30 g), fosfomycin (200 g), gentamicin (10 g), sulfamethoxazole-trimethoprim (25 g), and tetracycline (30 g). Isolates showing resistance to three or more antimicrobial drugs are considered multidrug resistant (MDR).
Whole-genome sequencing and comparative genomics. Eight of the isolates were sequenced, and comparative genomics analyses were done, including four strains from previous and ongoing studies, providing virulome, resistome, and whole-genome comparisons. Briefly, paired-end sequence data for eight strains were obtained using Illumina MiSeq for the following in silico analysis. The NGS QC Toolkit (v2.3.3) (36) was used to filter high-quality reads, followed by contig assembly using SPAdes Genome Assembler (v3.6.1) (37). Numbers of raw reads, respective read lengths, genome coverage obtained after filtering, and total number of contigs were computed. The generated de novo contigs were ordered and scaffolded using Contig-Layout-Authenticator (CLA) (38). Final draft genomes were obtained by merging scaffolds using a series of N's. Draft genomes were submitted to the RAST (39) server for annotation, and genome statistics from the resulting file were extracted using Artemis (40). The sequence type (ST) of each strain was determined by submitting contigs to https://cge.cbs.dtu.dk/services/MLST/.
Twelve strains, including eight from the current study and four from earlier reported and ongoing studies, belonging to different pathotypes were used for the following comparative analysis. BRIG (41) was used to visualize genome variation of the 12 strains. GeneMarkS (42) was used to predict protein sequences. BLASTp (43) analysis of the putative protein sequences was performed against the database of E. coli virulence genes downloaded from the Virulence Factors Database (VFDB) (44), providing a virulence profile for each strain. A specific virulence gene was considered present only if the BLASTp hit had identity greater than 60% and query coverage greater than 85%. The presence-absence status of all virulence genes carried by each strain was represented in the form of a heat map using R. A similar heat map was generated for putative resistance-related genes from the Comprehensive Antibiotic Resistance Database (CARD) (45), using BLASTp.
Phenotypic virulence determination. Biofilm formation was determined using M63 minimal medium as described earlier (13,46). Briefly, bacteria grown overnight were diluted to an optical density at 600 nm (OD 600 ) of 0.05 in M63 medium, and 200 l of each was inoculated in triplicate in a 96-well microtiter plate. OD 600 was measured, and the plate was covered with permeable sealing and incubated at 28°C for 48 h without shaking. Growth was measured after 48 h as OD 600 using a microtiter plate reader. Medium was removed gently, washed three times with deionized water, and dried. Bacteria were fixed with methanol (99%) for 15 min and stained with 1% crystal violet (30 min) after air drying. Plates were washed three times with deionized water and air dried again. The fixed stained bacteria were resolubilized using 200 l of ethanol-acetone (80:20), and absorbance was measured at 570 nm. Specific biofilm formation (SBF) was obtained using the formula SBF ϭ (AB Ϫ CW)/G, where AB is absorbance at 570 nm, CW is OD 570 of control well (without bacteria), and G is growth measured by the formula G ϭ OD 600(48 h) Ϫ OD 600(0 h) .
Serum bactericidal activity was assayed using 50% human serum as reported previously (13,46). Briefly, 5 l of overnight-grown culture was inoculated in 495 l of fresh LB broth and allowed to grow for 1.5 h at 37°C at 200 rpm. The bacterial cells were pelleted and resuspended in 1 ml of sterile 1ϫ phosphate-buffered saline (PBS). Thirty microliters of bacteria was added to 270 l of 50% human serum in a 96-well microtiter plate in triplicates, and 30 l of sample was taken out and plated on LB agar plates after dilution. The plate was covered and allowed to grow for 3 h at 37°C at 100 rpm. After 3 h of incubation, 30 l of sample was collected again from each well, diluted, and plated on LB agar plates. The bacteria were enumerated after overnight incubation at 37°C, and growth in serum was calculated by subtracting the CFU of 0 h from that of 3 h. A graph was plotted, and results were analyzed. Both of the assays were performed twice and in triplicate on a subset of 10 isolates from each category.
Statistical analysis. Statistical analysis of data for virulence and resistance genes was performed employing chi-square and Mann-Whitney tests for serum resistance and biofilm formation, respectively.