Novel Multiplex PCR Method and Genome Sequence-Based Analog for High-Resolution Subclonal Assignment and Characterization of Escherichia coli Sequence Type 131 Isolates.

Escherichia coli sequence type 131 (ST131) is a pandemic, multidrug-resistant extraintestinal pathogen. The multiple distinctive ST131 subclones differ for rfb and fliC alleles (O and H antigens), fimH allele (type-1 fimbriae adhesin), resistance phenotype and genotype, clinical correlates, and host predilection. Current PCR assays for detecting ST131 and its main subclones offer limited sub-ST characterization. Here we combined 22 novel and 14 published primers for a multiplex PCR assay to detect and extensively characterize ST131 isolates. The primers target mdh36, gyrB47, trpA72, sbmA, plsB, nupC, rmuC, kefC, ybbW, the O16 and O25b rfb variants, five fimH alleles (fimH22, fimH27, fimH30, fimH35, and fimH41), two fliC alleles (H4 and H5), a (subclone-specific) fluoroquinolone resistance-associated parC allele, and a (subclone-specific) prophage marker. The resulting amplicons resolve 15 molecular subsets within ST131, including 3 within clade A (H41 subclone), 5 within clade B (H22 subclone), and 7 within clade C (H30 subclone), which includes subclones C0 (H30S: 2 subsets), C1 and C1-M27 (H30R1: 2 subsets), and C2 (H30Rx: 3 subsets). Validation in three laboratories showed that this assay provides a rapid, accurate, and portable method for rapidly detecting and characterizing E. coli ST131 and its key subsets. Additionally, for users with whole genome sequencing (WGS) capability, we developed a command-line executable called ST131Typer, an in silico version of the extended multiplex PCR assay. Its accuracy was 87.8%, with most issues due to incomplete or fragmented input genome assemblies. These two novel assays should facilitate detailed ST131 subtyping using either endpoint PCR or WGS. IMPORTANCE These novel assays provide greater subclonal resolution and characterization of E. coli ST131 isolates than do the available comparable PCR assays, plus offer a novel sequence-based alternative to PCR. They may prove useful for molecular epidemiological studies, surveillance, and, potentially, clinical management.

INTRODUCTION

Escherichia coli sequence type 131 (ST131), which arguably is the most epidemiologically successful multidrug-resistant E. coli lineage ever identified, emerged dramatically beginning in approximately 2000 to become a pandemic pathogen (1–5). ST131 now accounts for the greatest share of multidrug-resistant human extraintestinal E. coli infections worldwide (6–12).

ST131 exhibits a complex clonal structure, as elucidated by phylogenetic analyses based on whole genome sequencing (WGS) data (2, 3, 5, 13) (Fig. 1). The most extensively expanded ST131 component is a recently evolved sub-lineage that has been designated variously as H30R—reflecting its association with both fluoroquinolone resistance and the type-1 fimbrial adhesin allele fimH30 (3)—and clade C, according to an alternate, trait-agnostic taxonomy (2, 5). Strains from H30R (clade C) typically exhibit the O25b somatic antigen and the H4 flagellar antigen.

FIG 1

Schematic of the Escherichia coli ST131 phylogeny. Branch lengths are for illustrative purposes only and are not proportional. Clades with undefined branching order are shown as polytomies. Main branches are labeled selectively as to sublcone or clade (using alternative taxonomies); terminal branches are labeled selectively as to O:H serotype, fimH allele, fluoroquinolone (FQ) phenotype (FQ-S [susceptible], FQ-R [resistant]), parC genotype, and PCR pattern no.

H30R in turn has important subcomponents, including subclones H30R1 (clade C1) and H30Rx (clade C2) (2, 3, 5, 13) (Fig. 1). Although most H30R1 strains are cephalosporin-susceptible, some – especially within an emerging subclade designated M1-27, which was identified first in Asia and is expanding now in Europe and North America – carry blaCTX-M-27 or, less commonly, blaCTX-M-14 (14–16). By contrast, most H30Rx (clade C2) strains carry blaCTX-M-15, the distinctive characteristic that led to ST131's initial discovery (3, 17).

In addition to H30R, H30 also has a basal, fluoroquinolone-susceptible H30S component (clade C0), which may be an evolutionary precursor to the pandemic H30R subclone (2, 3, 5, 13) (Fig. 1). The evolutionary steps and fitness factors that led from this epidemiologically unsuccessful ancestor to the world-changing H30R lineage remain incompletely understood, despite considerable study (2, 10).

Apart from the dominant H30 clade, other ST131 subcomponents also have distinctive resistance phenotypes, epidemiological associations, fimH alleles, O types, and fliC (flagellar H antigen) alleles (Fig. 1). For example, the ancestral H41 subclone (clade A) is associated with trimethoprim-sulfamethoxazole resistance, younger patients, fimH41, and the O16 and H5 antigens (18). Likewise, the phylogenetically intermediate H22 subclone (clade B), which typically is less extensively antimicrobial-resistant than the other ST131 subclones, is associated with food animals, including poultry and swine, where it can acquire antimicrobial resistance that subsequently transmits to humans (19). It is associated with fimH22 and, like the H30 subclone, the O25b and H4 antigens (20, 21). Like other ST131 subsets, the H22 subclone has a complex substructure, definitions of which are evolving.

ST131's clear clinical and epidemiological importance, and its various distinctive sublineages, has motivated the development of PCR-based assays for use in screening and characterizing clinical, commensal, and environmental E. coli isolates (18, 21–26). These assays have become increasingly sophisticated and discriminating over time, in response to emerging understandings of the ST131 substructure (23). Ultimately, WGS-based phylogenetic analysis and accessory trait identification provides the most granular and robust characterization of E. coli isolates (2, 10, 27–29). However, given the as-yet limited widespread availability and affordability of WGS technologies for general epidemiological and clinical use, we sought here to devise an enhanced multiplex PCR-based assay for the detection and molecular characterization of important ST131 sublineages, accessible to any laboratory equipped for conventional endpoint PCR. Additionally, for those researchers with access to WGS, we also developed a WGS-based method that aimed to replicate in silico the multiplex PCR-based assay.

RESULTS

Specificity for ST131.

In the J. Johnson Laboratory the PCR assay correctly identified as non-ST131 all 146 negative control isolates, i.e., for them it did not yield any of the three possible ST131-specific amplicons: mdh36, gyrB47, or trpA72. Thus, for ST131 per se the assay's estimated specificity was 100% (95% CI, 97.5–100%).

Subclade detection within ST131.

When the new assay was applied to the Price collection in the J. Johnson Laboratory, in comparison with the published WGS phylogeny (Supplemental material 1) it classified correctly by subclone all 105 ST131 strains (100% accuracy; 95% CI, 96.6% – 100%). Similarly, when it was applied to the Gordon collection in the D. Gordon laboratory, in comparison with an unpublished WGS phylogeny (Supplemental material 2) it classified correctly by subclone all 73 ST131 strains (100% accuracy; 95% CI, 95.1% – 100%).

In the J. Johnson Laboratory, the assay yielded 100% concordance with concurrently performed single or dual-target PCR assays for detecting the fimH30 allele, the H30Rx clade, and the O25b and O16 rfb variants. It also yielded 100% accuracy (95% CI, 96.6–100%) for detecting the parC E84V mutation, as identified by WGS (with or without supplemental Sanger sequencing), among 5 ST131 isolates with and 46 without this parC allele.

Assay stability.

In the J. Johnson Laboratory, after optimization the assay's performance was consistent across three different thermal cyclers: Eppendorf Mastercycler x50, Bio-Rad MyCycler, and Bio-Rad T1000. Notably, however, with the Bio-Rad T1000 a 1°C higher annealing temperature was required with all pools to avoid nonspecific amplification.

Diversity of PCR profiles.

Within the combined Price and Gordon collections (n = 179 isolates), the new assay identified 8 unique PCR profiles with just primer pool 1, whereas the addition of primer pools 2 and 3, and the 9-isolate Miscellaneous collection, increased this number to 15 (Table 1, Fig. 2). The 15 unique PCR profiles differed widely in frequency by collection (Table 2). The 8 most frequent profiles were encountered in both the Price and Gordon collections. Presence in only one—or neither—of these collections was limited to the 7 profiles with only one to three representatives each. These profiles included H30Rx:O-nontypeable (isolate JJ2449: Miscellaneous collection only), B0:fimH22 (isolate JJ1897: Price collection only), B1:fimH22-negative (isolates JJ2016, ZH71, and JMI025: Price collection only); H35Rx/C2 (isolates JJ2643 and U004: Price collection only); A:O25b:H4 (isolates M685551, M652483 and BS488: Gordon collection only), A:O25b:H5 (isolate M670745: Gordon collection only), and C0:parC E84V (isolate BS448: Gordon collection only).

TABLE 1

Distinguishing features of 15 unique PCR patterns observed among the present Escherichia coli study isolates with the novel 3-tube multiplex ST131 PCR assay

			ST131, subclones, clades										fimH					rfb		fliC		FQR
Pattern no.	ST131	Subclone/clade	gyrB47	mdh36	trpA72	nupC SNP	plsB SNP	kefC SNP	rmuC SNP	ybmW SNP	sbmA SNP	Prophage	fimH22	fimH27	fimH30	fimH35	fimH41	O16	O25b	H4	H5	parC E84V
1	+	H41/A	+	+	+	nab	nab	−	−	−	−	−	−	−	−	−	+	+	−	−	+	−
2	+	H41/A	+	+	+	nab	nab	−	−	−	−	−	−	−	−	−	+	−	+	+	−	−
3	+	H41/A	+	+	+	nab	nab	−	−	−	−	−	−	−	−	−	+	−	+	−	+	−
4	+	H22/B0	+	+	−	−	+	−	−	−	−	−	−	+	−	−	−	−	+	+	−	−
5	+	H22/B0	+	+	−	−	+	−	−	−	−	−	+	−	−	−	−	−	+	+	−	−
6	+	H22/B1	+	+	−	+	+	−	−	−	−	−	+	−	−	−	−	−	+	+	−	−
7	+	H22/B1c	+	+	−	+	+	−	−	−	−	−	−	−	−	−	−	−	+	+	−	−
8	+	H22/B1	+	+	−	+	+	−	−	−	−	−	−	−	+	−	−	−	+	+	−	−
9	+	H30S/C0	+	+	−	−	−	−	+	−	−	−	−	−	+	−	−	−	+	+	−	−
10	+	H30S/C0d	+	+	−	−	−		+	−	−	−	−	−	+	−	−	−	+	+	−	+
11	+	H30R1/C1, non-M27	+	+	−	−	−	+	+	−	−	−	−	−	+	−	−	−	+	+	−	+
12	+	H30R1/C1-M27	+	+	−	−	−	+	+	−	−	+	−	−	+	−	−	−	+	+	−	+
13	+	H30Rx/C2	+	+	−	−	−	−	+	+	+	−	−	−	+	−	−	−	+	+	−	+
14	+	H30Rx/C2e	+	+	−	−	−	−	+	+	+	−	−	−	−	+	−	−	+	+	−	+
15	+	H30Rx/C2f	+	+	−	−	−	−	+	+	+	−	−	−	+	−	−	−	+	+	−	+
n.a.b	−	non-ST131	+	−	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b
n.a.b	−	non-ST131	−	+	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b
n.a.b	−	non-ST131	−	−	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b	n.a.b

FQR, fluoroquinolone-resistant; SNP, single-nucleotide polymorphism; + and − signs, presence or absence of trait.

n.a., not applicable: isolates testing PCR positive for trpA72 (Pool 1) are assigned to clade O16/A and therefore results of PCR Pool 2 should not be considered as this pool is only designed to characterize the non-clade A types (clades B and C). (Although not applicable, clade A isolates do typically amplify the two clade B associated PCR amplicons in pool 2).

PCR Pattern 7 is representative of a fimH allele type (fimH94) not detected by primers in this PCR assay.

PCR Profile 10 retains the historical consensus subclade label “H30S/CO” as determined by kefC clade C PCR positive results and rmuC (C1) and ybbW (C2) negative PCR result. The parC E84V positive PCR and corresponding fluoroquinolone-resistant phenotype differentiate this profile type from Profile 9, representing the consensus H30S/C0 representative being parC E84V negative with a fluoroquinolone-susceptible phenotype.

fimH35 variant of H30Rx/C2 subclade.

O-type negative (as determined by classic serotype methods at the E. coli Reference Center (The Pennsylvania State University), and also reflected within assay, PCR negative for bothO25b and O16 PCR negative variant of H30Rx/C2 subclade).

FIG 2

PCR profiles for primer pools 1–3 with representative ST131 and non-ST131 Escherichia coli isolates. Primer pools are listed to right of gel images (pool 1, top; pool 2, middle; pool 3, bottom). Lanes 1–14, representative ST131 isolates for 14 of the 15 PCR patterns observed with the new assay, as listed below image. (The missing pattern was observed only in the D. Gordon laboratory; the representative isolates was unavailable in the J. Johnson laboratory for inclusion in this image.) Lane L, 100 bp ladder. Lanes 15–19, diverse non-ST131 E. coli isolates from STs (lane no.) 10 (15), 73 (16), 95 (17), 144 (18), and 69 (19). Horizontal text boxes above gels list, from top to bottom, ST131 status; ST131 clade/subclade; O type; H type; and fimH allele. Labels to left of gels identify each band and its size in base pairs (bp). Phylogenetic subsets are labeled using alternate taxonomies.

TABLE 2

Multiplex PCR profile types, associated traits, representative strains, and number per study collection

					Collection
PCR type	Clade	O:H type	fimH allele	Representative strain	Price (n = 105)	Gordon (n = 73)	MVAST (n = 20/90)a
1	A	O16:H5	41	MVAST020	7	11	5
2	A	O25b:H4	41	BS488	0	3	0
3	A	O25b:H5	41	M670745	0	1	0
4	B0	O25b:H4	27	H17	1	1	0
5	B0	O25b:H4	22	JJ1897	2	0	0
6	B1	O25b:H4	22	JJ1969	28	21	1
7b	B1	O25b:H4	94b	ZH071b	3	0	0
8	B1	O25b:H4	30	G199	1	3	0
9	C0	O25b:H4	30	CD306	4	2	0
10c	C0c	O25b:H4	30	BS448c	0	1	0
11d	C1	O25b:H4	30	JJ2193	32	11d	10
12	C1-M27	O25b:H4	30	U024	1	7	1
13	C2	O25b:H4	30	JJ1886	24	12	3
14	C2	O25b:H4	35	U004	2	0	0
15e	C2	ONTe:H4	30	JJ2449e	0e	0e	0e

From the (n = 90) all-comers MVAST collection (Table Y) only the 20 ST131 isolates are shown here.

The fimH94 allele is not part of the fimH22 allele complex and is not detected by this multiplex PCR assay. Thus, PCR profile type #7 lacks any fimH allele band.

Isolate BS448, although from clade C0, has the typical gyrA/parC haplotype of fluoroquinolone-resistant H30R (C1, C2) isolates, and with primer pool 1 produces the expected parC amplicon.

Two of the 11 isolates that represent PCR profile type #11 have fimH99 (which exhibits 1 single-nucleotide polymorphism difference versus the fimH30 allele, not in the primer region).

ONT, O-nontypeable. This profile uniquely lacks both the O25b and O16 rfb amplicons. The only known representative isolate, JJ2449, is from the “Misc.” collection (Table 4), not one of the collections listed here.

TABLE 4

Strain sets used for assay derivation, validation, and portability assessment

Name	No. isolates	Nature of collection	Available metadata	How used	Which lab(s)	Ref.
Price	105	Diverse ST131 isolates	WGS, phylogeny, accessory traits	Derivation, validation	JJ	(3)
Gordon	73	Diverse ST131 isolates	WGS, phylogeny, accessory traits	Derivation, validation	DG	n.a.
MVAST	90	Consecutive clinical isolates	Single-target PCR results	Validation for clinical use	JJ	n.a.
ECOR	72	Reference E. coli isolates	WGS, fimH allele, ST	Non-ST131 neg. controls	JJ	(33)
BUTI	66 (65b)	Historical urosepsis isolates	WGS, fimH allele, ST	Non-ST131 neg. controls	JJ	(34, 35)
Misc.	9c	Extra STs, PCR type, phylogroup	WGS and/or ST; O:H serotype	ST131/non-ST131 controls	JJ	n.a.
Test	19d	Representative type strains	Novel PCR assay results	Portability assessment	JJ, TJ	n.a.

lab, laboratory; Ref., reference; WGS, whole genome sequencing; JJ, James Johnson; DG, David Gordon; ST, sequence type; TJ, Timothy Johnson; n.a., not applicable (i.e., unpublished or multiple sources); Misc., miscellaneous.

This 66-isolate collection contains one ST131 isolate and 65 non-ST131 isolates.

Includes 8 non-ST131 isolates, from phylogroups A (ST617), B2 (ST1193), D (ST38), E (ST11), F (ST354), and G (ST 117, ST174), and 1 ST131 isolate (clade C2; fimH30; non-O25b/non-O16).

Selected deliberately as 12 type-representative ST131 isolates and 7 phylogenetically diverse non-ST131 isolates, from phylogroups A (ST10), B2 (ST12, ST73, ST95, ST144, ST1194), and D (ST69).

Comparison with established PCR assays among fresh clinical E. coli isolates.

To simulate a clinical diagnostics application, 90 consecutive fresh clinical isolates from the MVAMC microbiology laboratory were tested in parallel using (i) the novel assay and (ii) established single or dual-target PCR assays. The assays were compared for their results regarding ST131 status, fimH30, H30Rx, and the O25b and O16 rfb variants. The compared assays yielded 100% concordant results (95% CI, 96.4% - 100%), jointly identifying as ST131 the same 20 isolates, of which 5 were O16/A (all non-H30) and 15 were O25b (1 non-H30/B and 14 H30/C, including 3 H30Rx/C1). This collection's 20 ST131 isolates exhibited the 5 most common ST131-associated PCR patterns encountered in the Price and Gordon collections (Table 2).

Additionally, in this testing of fresh clinical isolates the new multiplex assay augmented the conventional assays in several ways. Specifically, it confirmed the 5 O16/A ST131 isolates by both traditional rfb O16 and the trpA clade A-specific amplicon (negative in all O25 ST131 isolates) and confirmed the 3 H30Rx/C2 isolates based on clade-specific SNPs in both sbmA and ybbW. It also uniquely assigned one of the 14 H30/C isolates to the (recently described) C1-M27 clade, assigned the single non-H30 O25b isolate to the B1 subclade, showed that this isolate had the fimH22 allele, identified all O16 isolates as having fliC H5 and fimH41, and identified all O25 isolates as having fliC H4.

Portability assessment.

For a formal assessment of the new assay's portability across laboratories it underwent independent, blinded testing in the T. Johnson Laboratory, which is geographically removed from the originating (J. Johnson) laboratory. After an initial optimization phase involving known positive and negative controls, plus open between-laboratory communication regarding technical issues, blinded validation testing was done using a panel of 18 coded test isolates (12 ST131, 6 non-ST131) (Supplemental material 3). The blinded testing correctly identified the isolates' ST131 status, but initially yielded discrepancies regarding their extended PCR profiles. Assay modifications were implemented to address these discrepancies, including (i) length adjustments in one fimH22 primer and one ybbW primer and (ii) more explicit instructions for gel interpretation. These modifications yielded 100% accurate assay results, without a need to adjust annealing temperatures.

ST131Typer performance assessment.

The overall accuracy of the (WGS-based) ST131Typer as a surrogate for the PCR assay, based on the 255 isolates tested, was 87.8% (95% CI, 83.2–91.6%). However, performance varied greatly by endpoint. Specifically, ST131Typer correctly identified all 78 negative control isolates as non-ST131 (100% specificity; 95% CI, 95.4–100%) and correctly classified 171 of the 173 ST131 strains by subclade (97.7% accuracy; 95% CI, 94.3–99.4%). By contrast, ST131Typer yielded the same PCR profile as the in vitro extended multiplex PCR assay for only 142 (82.5%; 95% CI, 76.1–87.8%) of the 173 ST131 strains.

Analysis of the 31 total instances where ST131Typer results did not match the PCR profile showed that 19% (6/31) were due to ST131Typer profile misclassifications and 81% (25/31) to unknown profile classifications. Supplemental material 5 lists the strains with a discrepancy, their placement in the phylogeny, their associated characteristics, and the specific discrepancies. (For the remaining 142 strains, ST131Typer and PCR agreed for all markers and classifications.) All 31 ST131Typer errors (one per strain, for 31 strains) were caused by inability to identify a single required target sequence, specifically the rfb O25b variant (16 strains), fliC H4 (7 strains), kefC (4 strains), and fimH22, fimH30, fimH40, and prophage marker (1 strain each).

Potentially explaining nearly all missing rfb O25b targets (15/16), and all four missing kefC targets, were highly fragmented and incomplete genome assemblies, with the forward and reverse primer sequences on different contigs, or one or both not detected. Indeed, N50 values (a common metric for genome assemblies; smaller values indicate greater fragmentation) were considerably lower for assemblies with incorrect, versus correct, PCR profile classifications (Mean ± SD: 58,963 ± 57,937 versus 224,646 ± 564,985). When use as a supplemental typing method CGE’s SerotypeFinder identified the missing rfb O25b variant in 15 of the 16 strains where ST131Typer missed it. However, match identities were <100%, and high-scoring BLAST alignments typically shorter than the reference rfb O25b sequence.

Additional instances of mismatch between ST131Typer and the PCR assay were potentially from sequence variation in the forward or reverse primer regions. This was particularly evident with the missing fliC H4 and fimH target sequences, where the primer regions frequently included one or more SNPs. As for the rfb O25b target, SerotypeFinder identified the fliC H4 variant in all these samples, but with match identities <100% and high-scoring BLAST alignments shorter than the reference fliC H4 sequence. Likewise, for the three samples with a missing fimH allele per ST131Typer, CGE’s FimTyper identified fimH alleles that represented minor (1–2 nucleotide) variants of the alleles detected by PCR.

DISCUSSION

In this study, we developed and extensively validated a novel 3-pool, multiplex PCR assay—and a WGS-based, in silico analog of it—for identifying and characterizing E. coli ST131 isolates. The PCR assay demonstrated an excellent ability to distinguish E. coli ST131 from other E. coli, to classify ST131 isolates by subclonal type (in comparison with phylogenomic analyses and established PCR-based assays), and to characterize the isolates for key accessory traits, including relevant fimH and fliC alleles, the O25b versus O16 rfb variants, and a fluoroquinolone resistance-associated parC SNP. The assay also was portable across laboratories and performed well with fresh clinical isolates. Thus, this novel PCR assay should prove useful for molecular epidemiological studies and, potentially, clinical diagnostics. The in silico version of the assay, ST131Typer, accurately distinguished E. coli ST131 from other E. coli, and classified ST131 isolates accurately into subclades. However, its ability to replicate the PCR assay's full profile patterns was impeded by incomplete genome assemblies and minor sequence variation.

Several PCR-based assays have been described previously—including by some of the present authors—for detecting ST131 per se, key ST131 subsets, and various relevant accessory traits (18, 21–26). The most extensive of these assays differentiates ST131 from non-ST131 isolates; resolves ST131 clades A, B, and C; and, within clade C, resolves subclades C1-non-M27, C1-M27, C2, and “C, not C1 or C2” (23). Additionally, in that assay the presence of an ST131 amplicon but absence of all clade/subclade amplicons identifies an intermediate branch between clades B and C. The present novel assay supersedes that and other previous assays by detecting multiple subsets and identifying relevant accessory traits within all clades and subclades.

The present multiplex assay, with its extensive range of targets, yielded several novel findings. Specifically, within clade A it identified uncharacteristic O25b-associated variants, in addition to the expected O16-associated variants, and separated these according to fliC allele. Comparison with a SNP-based phylogeny (Supplemental material 2) showed that this variation in accessory traits within clade A was phylogenetically distributed, with the “atypical” variants occupying a basal position. This suggested that the O25b rfb allele and fliC H4 were possibly ancestral within clade A and, hence, perhaps within ST131 as a whole.

Additionally, within clade C2 the new assay differentiated isolates with (uncharacteristic) fimH35 from those with (characteristic) fimH30. Likewise, within subclades B0 and B1 it differentiated isolates with (uncharacteristic) fimH27 from those with (characteristic) fimH22, or neither of these. It also identified within clade C0 an isolate (BS448) that uniquely exhibited the distinctive fluoroquinolone resistance-associated parC SNP (a251t = E84V) that typifies clades C1 and C2, that is, H30R (3, 30). Notably, according to WGS analysis (not shown) that isolate also had the three other H30R-associated fluoroquinolone resistance-conferring replacement mutations in gyrA (c248t = S83L; g259Aa = D87N) and parC (g239t = S80I) (30), suggesting that it may be an immediate ancestor to the H30R pandemic lineage. Finally, absence of the O25b and O16 rfb variants in one H30Rx/C2 isolate defined the isolate as O-nontypeable by the present assay; its O-nontypability was confirmed at a reference laboratory (E. coli Reference Center, The Pennsylvania State University) by conventional broad-range O serotyping. Whether this isolate's distinctive PCR profile represents a new O-negative H30Rx subtype or is unique to this particular isolate remains to be determined.

Although the novel assay was designed around three primer pools (i.e., three PCRs), the individual pools can be used separately, including in a hierarchical manner. For example, generic E. coli isolates can be screened initially using pool 1, which in addition to detecting ST131 per se can resolve 8 different subclonal profiles. Isolates so identified as ST131 can then be further characterized using pools 2 and 3, which offer refined subclone/confirmatory discrimination and can identify additional accessory traits, including H30R's hallmark fluoroquinolone resistance-associated parC SNP (3, 30).

When validated using an extensive panel of positive and negative control strains, the assay yielded 100% accuracy in comparison with external reference standards, including WGS analyses or individual PCR results. Additionally, it was portable across laboratories. Notably, however, in each new laboratory some optimization was needed. Thus, when the assay is established in additional laboratories further optimization may be needed to ensure the expected performance with relevant control strains.

The assay has notable limitations. First, other than the hallmark fluoroquinolone resistance-associated mutation in parC, it lacks resistance gene targets, including the various ST131-associated blaCTX-M variants and carbapenemase-encoding genes, for which supplemental assays would be needed. Second, its amplicon profiles are complex, potentially posing interpretive challenges. Adherence to detailed instructions and use of the provided interactive data entry spreadsheet can help. Third, the particular sublineages and associated traits included in the assay represent only a fraction of the ever-expanding number of possible targets and were selected somewhat arbitrarily. For example, the assay addresses minimally the recently appreciated clonal diversity within the H22/clade B lineage (19, 20). The present version of the assay represents a compromise between comprehensiveness and practicality; future iterations could address emerging markers of interest. Fourth, assay optimization will be needed in new user laboratories, possibly obliging communication with the developers.

Due to the increasing demand for sequence-based analytical tools, we also developed ST131Typer as a WGS-based alternative to the in vitro PCR assay. ST131Typer was highly accurate in distinguishing E. coli ST131 isolates from other E. coli and in classifying them into ST131 subclades but was only moderately accurate in replicating the PCR assay's profiles. The main issue appeared to be fragmented or incomplete input assemblies, which, in turn, were likely caused by poor quality reads and/or insufficient genome sequence coverage. Thus, the accuracy of ST131Typer's results depends on sequence and assembly quality. Currently, sufficient next-generation sequencing coverage and high-quality genome assembly are still major challenges in some research and clinical settings. Analysis of WGS data often requires considerable computational resources and a bioinformatics skill set, and associated costs can be prohibitive if large numbers of samples are processed. As such, the novel in vitro PCR assay presented here may be a useful, even preferable, ST131 characterization tool for general epidemiological and clinical use, especially in resource-limited settings.

In conclusion, this novel PCR assay, consisting of 3 multiplex primer pools, can rapidly and specifically distinguish ST131 from other E. coli and characterize ST131 isolates according to multiple phylogenomic subclones and accessory trait, yielding 15 distinct trait profiles with the present reference strains. ST131Typer, a novel web-accessible, WGS-based tool, provided a reasonably accurate in silico surrogate for the PCR assay, especially when supplemented with CGE-supported subroutines. Pending the wider availability, affordability, and user-friendliness of WGS and the associated bioinformatics approaches, this PCR assay should prove useful for epidemiological studies and, potentially, clinical diagnostics.

MATERIALS AND METHODS

Assay design and primers.

The assay was designed to identify two different types of entities: phylogenetic subsets (seven total) and accessory traits (four distinct traits; 10 total variants thereof). The seven phylogenetic subsets of interest (alternate taxonomy labels in parentheses) included the three major ST131 subdivisions, i.e., subclones H41 (clade A), H22 (clade B), and H30 (clade C), and components thereof. These components were, within the H22 subclone, subclade B0 (versus the rest of subclade B, hereafter referred to as B1); within the H30 subclone, subclones H30S (C0), H30R1 (C1), and H30Rx (C2); and within subclone H30R1, the (blaCTX-M-27-associated) C1-M27 subclone. The accessory traits of interest included diverse ST131-associated variants of four key genes: fimH (type-1 fimbriae adhesin) (31), fliC (flagellar H antigen), rfb (O antigen), and parC (topoisomerase), specifically the (H30R1 and H30Rx)-associated E84V parC variant that is associated with higher fluoroquinolone MICs than those of other fluoroquinolone-resistant strains (30, 32).

Relevant published primers were used as-is or were modified as needed for optimized performance and compatibility with other primers. Novel primers were designed based on published target sequences and were optimized – for sensitivity, specificity, amplicon size, and compatibility with other primers – based on in vitro performance against relevant positive and negative control strains, both as individual primer pairs and in multiplex pools. Amplicon specificity was assessed by comparing the observed product size with the expected product size for the particular target, not by sequencing. Primer sensitivity and specificity were assessed based on ability to amplify a product of the expected size with positive control strains but not negative control strains.

The final assay comprised 36 total primers: 22 were novel and 14 were previously published (Table 3). These primers target allele-specific single-nucleotide polymorphisms (SNPs) or nucleotide sequences in mdh36 and gyrB47 (ST131-specific), the O16 and O25b rfb variants, 5 fimH alleles, 2 fliC alleles, the fluoroquinolone resistance (FQR)-determining region of parC and selected other clade-specific genes or alleles thereof. The subclade markers (associated target clade/subclade) include trpA72 (clade A), plsB (clade B1), nupC (clade B, including subclades B1 and B0), kefC (clade C, including subclades C0, C1, and C2), rmuC (subclade C1), ybbW and sbmA (subclade C2), and a prophage marker (C1-M27 subclade).

TABLE 3

Primers targets and primer sequences for the three multiplex pools

Pool	Target	Target significanceb	Product size (bp)	Forward (for) and reverse (rev)primer sequence (5′–3′)	Conc. (uM)	Reference
1	Prophage	C1-M27 clade	822	(For) TTACTCCGACTATGCGTTCAC	0.25	(16)
				(Rev) CTTCTTCGCTGGCAACTTCT	0.25	This study
	rfb O16 allele	O16 antigen	732	(For) GAAGGTGTTACCACCATCAATG	0.25	This study
				(Rev) CAGGATCATTTATGCTGGTACG	0.25	(44)
	rfb O25b allele	O25b antigen	557	Same as O16 forward primer	n.a.c	This study
				(Rev) ATGCTATTCATTATGCGCAGC	0.25	(44)
	trpA72	O16 subclone/clade A	487	(For) CTTTTGCGGCGGGAGTA	0.25	This study
				(Rev) GAAATCGCGCCCGCAGT	0.25	(18), this study
	fimH30 allele	fimH30 allele complex	350	(For) GCCAATGGTACCGCTATT	0.25	(30)
				(Rev) GCTTTAATCGCCACCCCA	0.25	(30)
	mdh36	ST131-specific SNP	273	(For) TTAACGTTAACGCCGGT	0.4	(22)
				(Rev) TGGTAACACCAGAGTGACCA	0.4	(22)
	gyrB47	ST131-specific SNP	138	(For) CTGCGCGATAAGCGCGAC	0.15	(22)
				(Rev) AATACCGTCTTTTTCGGTGGAA	0.15	(22)
	sbmA	H30Rx subclone/clade C2	64	(For) GCTGCCTACCGTAAAGAA	0.25	This study
				(Rev) CCGTAGGCGGCGTCA	0.25	This study
2	plsB	H22 subclone/clade B1	628	(For) AGAATTCCGGCGCGTTG	0.4	This study
				(Rev) CGCGAGCAGTTAACCG	0.4	This study
	nupC	H22 subclone/clades B0, B1	498	(For) TCCCGCAATCGTATGTACACC	0.4	This study
				(Rev) ATCCATCATCGCAACGAAC	0.4	This study
	rmuC	H30R1 subclone/clade C1	353	(For) TATCGCGTCGGTGCGTAAC	0.25	This study
				(Rev) CCTGCGCTTTGTCGAGAT	0.25	This study
	kefC	H30 subclone/clade C (C0,C1,C2)	238	(For) CGCAAATGGGCCGCAGA	0.25	This study
				(Rev) GCACTAAACACTTCCCGCA	0.25	This study
	ybbW	H30Rx subclone/clade C2	188	(For) GTTGCGGTCTGGGCA	0.1	This study
				(For) TCCAGCACGTTCCAGGTG	0.1	This study
	parC	E84V amino acid substitution	107	(For) GGCGATATTGCCTGTTATGT	0.25	This study
				(Rev) TTCGGATCGTCCGGCGCC	0.25	This study
3	fliC H5 allele	H5 flagellar antigen	614	(For) CGTTCAAGCGACGACCGGG	0.4	This study
				(Rev) ATGTTACATATCCCGCAGCCGA	0.4	This study
	fimH35 allele	fimH35 allele complex	500	(For) CTGTAAAACCGCCAATGGTACA	0.4	This study
				(Rev) TGACATCACGAGCAGAAACATCG	0.4	This study
	fimH27 allele	fimH27 allele complex	417	(Same as fimH22 forward)	n.a.c	This study
				(Same as fimH35 rev)	n.a.c	This study
	fimH22 allele	fimH22 allele complex	279	(For) TGGGGCAAAACCTGGTCGTG	0.2	This study
				(Rev) CAGCTTTAATCGCCACTCCC	0.2	This study
	fliC H4 allele	H4 flagellar antigen	199	(For) GGCGAAACTGACGGCTGCTG	0.08	(45)
				(Rev) ACCAACAGTTACCGCCGC	0.08	(45), this study
	fimH41 allele	fimH41 complex	92	Same as fimH35 forward	n.a.c	This study
				(Rev) TTTGCCCCACATTCACGGCG	0.2	This study

For, forward; Rev, reverse; Conc., concentration.

Phylogenetic subsets are labeled according to both the Price et al. taxonomy (which uses the term “subclone”) ref and the Petty et al. taxonomy (which uses the term “clade”) ref. In fimH-based subclone designations the H is italicized, to imply “fimH,” and avoid confusion with flagellar H antigens.

n.a., not applicable: shared primers are used in combination with unique primers to amplify different PCR targets. The shared primer should be added only once, in the concentration first specified.

The primers were split into three multiplex reaction pools based on band size and annealing temperature (Table 3). Pool 1 included primers specific for ST131 per se; its subclades H41-O16 (clade A), H30Rx (clade C2), and C1-M27; the fimH30 allele; and the O16 and O25b rfb variants. Pool 2 included primers specific for subclades H22 (clade B), B1, H30 (clade C), H30R1 (clade C1), and H30Rx (clade C2: using a different target than pool 1), plus the E84V parC mutation. Pool 3 included primers specific for the H4 and H5 fliC alleles and the fimH22, fimH35, and fimH41 allele complexes, with allele complex defined here as the type allele plus minor variants thereof.

PCR conditions.

The three primer pools were used in separate PCRs and were tested on different thermal cyclers. A hot-start Taq polymerase was used (GoTaq Hotstart, Promega) with the supplied buffer, adjusted to contain either 4 mM MgCl2 (pools 1 and 2) or 2 mM MgCl2 (pool 3.) The cycle routines were as follows: pool 1, 95°C 2 min; 30 cycles each (94°C 15 s, 65°C 1 min 15 s), 72°C 2 min; pool 2, 95°C 2 min; 34 cycles each (94°C 15 s, 67°C 45 s), 72°C 2 min; and pool 3, 95°C 2 min; 38 cycles each (94°C 15 s, 1 min 15 s 65°C), 72°C 2 min. Reactions were then held at 12°C. Amplicon bands were resolved by 2% agarose gel electrophoresis and ethidium bromide staining.

Strain sets and phylogenetic analysis.

Seven strain sets (432 total isolates) were used for assay derivation, validation, and portability assessment (Table 4). The individual strain sets and how they were used here are described below.

Non-ST131 controls.

The negative controls were 146 phylogenetically diverse non-ST131 E. coli isolates, including 138 from two published collections with available WGS data and eight supplemental isolates (assigned to the Miscellaneous collection) representing additional STs not found in these two collections (Table 4). The two published collections included the E. coli Reference (ECOR) collection (n = 72) (33) and a set of bloodstream isolates from patients with bacteremic urinary tract infection (n = 66 isolates) (34, 35). The eight supplemental isolates were from diverse-source collections in the J. Johnson Laboratory. Collectively, the negative controls included all major E. coli phylogroups (A, B1, B2, C, D, E, F, and G) (36, 37), each represented by from 3 to 30 ST-fimH types (31) (Table 5), giving 97 total unique non-ST131 ST-fimH types (Table 5).

TABLE 5

Distribution of non-ST131 Escherichia isolates (n = 146) by phylogenetic group, ST, and ST-fimH allele combination

Phylo-groupa	Isolates (no.)	Unique STsb (no.)	Unique STb-fimH allele combinationsc (no.)	Specific STs included
A	29	16	24	10, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 63, 65, 77, 87, 617
B1	16	13	13	53, 54, 55, 56, 58, 67, 75, 84, 85, 86, 88, 89, 101
B2	60	19	30	12, 14, 73, 74, 76, 78, 79, 80, 81, 82, 83, 95, 127, 144, 550, 625, 998, 1193, 4862
C	8	3	3	23, 88, 410
D	13	12	13	38, 66, 68, 69, 70, 71, 72, 106, 393, 405, 501, 598
E	9	4	5	11, 57, 61, 64
F	9	6	7	59, 60, 62, 354, 618, 648
G	2	2	2	117, 174
Any	146	75	97	All of above

Phylogroup defined by Clermont Typer (in silico, using whole genome sequencing data) (46) or multiplex PCR (47).

STs, sequence types, as defined using Achtman multilocus sequence typing (MLST) (https://enterobase.warwick.ac.uk/species/ecoli/allele_st_search).

ST-fimH allele combinations provide sub-ST discrimination.

ST131 controls.

The positive controls were 178 diverse genome-sequenced ST131 isolates (Table 4). These included 105 isolates from Price et al. (3) (hereafter, the Price collection), 73 isolates from the D. Gordon laboratory (37) (hereafter, the Gordon collection), and one isolate with a unique ONT:H4 serotype (assigned to the Miscellaneous collection). The Price and Gordon collections, as described in greater detail below, were selected deliberately from larger private collections of ST131 isolates to provide a broad range of phylogenetic subsets, sources, and accessory trait profiles.

The Price collection (n = 105) was used in the first WGS-based phylogenetic analysis of ST131 (3) and was the main component of subsequent studies that extended this analysis with additional strains and bioinformatics approaches (2, 10, 13). These analyses largely concurred regarding the sub-ST clonal structure of ST131, but used alternate taxonomies for the observed cladal subsets, one based on clade-associated accessory traits (i.e., fimH alleles, O types, and antimicrobial resistance), the other on a (trait-agnostic) alpha-numeric system. According to the latter taxonomy, the Price collection is distributed as follows (no. isolates per clade or subclade): 7 A, 3 B0, 32 B1, 4 C0, 32 C1, 1 C1-M27, and 26 C2 (Supplemental material 1).

The Gordon collection (n = 73), which represents diverse sources, was distributed as follows in an unpublished core genome, SNP-based phylogeny (no. isolates, clade or subclade [per alpha-numeric taxonomy]): 15 A, 1 B0, 24 B1, 3 C0, 11 C1-non-M27, 7 C1-M27, and 12 C2 (Supplemental material 2). For this phylogeny, the SNP data were derived using PARSNP (https://github.com/marbl/parsnp), with recombination detection/filtration, after which RAxML and PhyML trees were generated using TOPALi v.2 (http://www.topali.org/). Sequence-based results for this collection also were available for predicted O:H serotype, fimH allele, and fluoroquinolone resistance-associated gyrA/parC mutations. These isolates underwent PCR-based testing in the Gordon laboratory, using the present novel assay according to protocols provided by the (source) J. Johnson Laboratory.

For the Price and Gordon collections, subclone/clade assignments (which were used as the reference standard for assessing the multiplex PCR assay) were based primarily on isolates' position within the corresponding core genome SNP-based phylogeny, for which subclone/clade identity was apparent from (i) comparisons with published ST131 phylograms and (ii) the consensus characteristics of a given subclone or clade's constituent isolates. For the several strains that were placed ambiguously in the phylograms, i.e., at the junction of the H22 and H30 subclones (clades B and C) or of the H30S and H30R subclones (C0 and C1/C2), subclone/clade assignment was resolved based on presence/absence of subclone/clade-specific genomic markers, as detected by WGS. Specifically, membership in the H30S subclone (clade C0) was defined by placement in the phylogeny within or immediately basal to the H30 subclone (clade C), but basal to the H30R1 and H30Rx subclones (clades C1 and C2), plus (according to WGS) containing the inclusive H30 subclone (clade C) marker kefC but lacking the markers for subclones H22 (clade B: nupC), H30R1 (C1: rmuC), and H30Rx (C2: ybbW and sbmA) (Table 3).

Fresh clinical isolates.

Assay performance also was assessed in real time, using fresh clinical isolates. For this, 90 consecutive clinical E. coli isolates were retrieved over a 3-month span from the Minneapolis VA Medical Center clinical microbiology laboratory and tested in parallel using both the novel assay and established single-plex PCR assays for the detection of ST131, the O25b and O16 rfb variants, the fimH30 allele, and the H30Rx/C2 subclone (38).

Independent assay validation and blinded testing.

For a formal validation of the novel PCR assay in a laboratory other than that in which it was developed, an unmasked set of ST131 and non-ST131 E. coli isolates was provided to the T. Johnson Laboratory (University of Minnesota), along with the recommended assay protocol. In the T. Johnson Laboratory, the assay was done using a similar DNA extraction method (boiled lysates of overnight broth-grown bacteria) and similar reagents (including GoTaq Hotstart, Promega) as in the J. Johnson laboratory, but using a different type of thermal cycler (MyCycler, Bio-Rad, versus Mastercycler x50, Eppendorf). PCR products were separated in 2% agarose gels with ethidium bromide for visualization with UV light. Annealing temperatures and cycle numbers were adjusted as needed to achieve amplicon profiles for each pool similar to those shown in the provided protocol.

Once the assay was established in the T. Johnson Laboratory, a test set of 12 randomly assorted unique ST131 and 7 non-ST131 isolates (Supplemental material 3) was analyzed in a blinded manner. Results from the 3 multiplex pools were assessed for presence/absence of amplicons of the expected size for the particular pool. Data were entered into an interactive Excel file containing integrated interpretive algorithms (provided by the J. Johnson Laboratory: Supplemental material 4), which generated presumptive ST131 subclade assignments based on the binary amplicon data.

ST131Typer description and performance assessment.

To create an in silico version of the extended multiplex PCR assay that utilizes WGS data, a command-line script was written to extract target sequences via the 20 primer pairs described in Table 3. Because the goal was to replicate in silico the in vitro multiplex PCR assay, and because the multiplex PCR assay infers amplicon specificity based strictly on amplicon size, the in silico version does so as well. To run an analysis, the user calls the bash script ST131Typer.sh, which accepts two mandatory arguments: a single FASTA/multi-FASTA assembly file or a directory of multiple FASTA assembly files, and the user-defined output directory. Amplicons are then identified in each input assembly using the SeqKit software’s seqkit amplicon function (39) and the provided tab-separated file of primer pair sequences. A valid amplicon is identified only if (i) 100% sequence identity is found between the primers and the input assembly sequence and (ii) both forward and reverse primer sequences are found on the same contig. No sequence analysis is done of the amplicon except for the primer-binding regions. Amplicon sequences greater than or less than 5% of the expected amplicon size are flagged in the final report with asterisks.

The ST131 PCR profile is then determined based on the same presence/absence patterns of subclone/clade-specific genomic markers as used in the in vitro extended multiplex PCR assay. If the target sequence for either mdh36 or gyrB47 is missing, the assembly is flagged as “non-ST131” and the remainder of the presence/absence amplicon pattern is ignored. If the mdh36 and gyrB47 targets are both present, but the resulting presence/absence amplicon pattern is not one of the PCR profiles described in Results, the assembly is characterized as unknown. If O-type, H-type, and/or fimH allele cannot be determined, the final report includes a warning message and recommends using additional typing tools such as the Center for Genomic Epidemiology’s (CGE's) SerotypeFinder (40) and FimTyper (41). The single output directory includes both a tab-separated final report summarizing (i) the results of all input assemblies and, for each assembly, (ii) the seqkit amplicon output, in BED format. The ST131Typer script, associated data files, and user instructions are freely available at the following GitHub repository: https://github.com/JohnsonSingerLab/ST131Typer.

To assess the accuracy of ST131Typer, all isolates used for the extended multiplex PCR assay derivation and validation, and for which WGS data were publicly available, were used (n = 255). This included 78 non-ST131 E. coli isolates (primarily from the ECOR collection) and 177 phylogenetically diverse ST131 isolates (primarily from the Price and Gordon collections). Assemblies were available in EnteroBase (42) for most isolates. For the remaining isolates, Illumina paired-end reads were downloaded from NCBI’s Short Read Archive database and processed using the Shovill pipeline with the SPAdes assembler, adaptor trimming enabled, and genome size set to 5,000,000 bp (https://github.com/tseemann/shovill). Assembly quality was assessed using QUAST (43) and the EC958 ST131 reference genome GCF_000285655.3.

Statistical methods.

PCR assay and ST131Typer performances were assessed as both specificity and overall accuracy, in comparison with relevant controls. Reported results included the point estimate and 95% confidence interval.

Data availability.

ST131Typer is freely available and accessible at https://github.com/JohnsonSingerLab/ST131Typer.