E. coli Common pili promote the fitness and virulence of a hybrid aEPEC/ExPEC strain within diverse host environments

ABSTRACT

Pathogenic subsets of Escherichia coli include diarrheagenic (DEC) strains that cause disease within the gut and extraintestinal pathogenic E. coli (ExPEC) strains that are linked with urinary tract infections, bacteremia, and other infections outside of intestinal tract. Among DEC strains is an emergent pathotype known as atypical enteropathogenic E. coli (aEPEC), which can cause severe diarrhea. Recent sequencing efforts revealed that some E. coli strains possess genetic features that are characteristic of both DEC and ExPEC isolates. BA1250 is a newly reclassified hybrid strain with characteristics of aEPEC and ExPEC. This strain was isolated from a child with diarrhea, but its genetic features indicate that it might have the capacity to cause disease at extraintestinal sites. The spectrum of adhesins encoded by hybrid strains like BA1250 are expected to be especially important in facilitating colonization of diverse niches. E. coli common pilus (ECP) is an adhesin expressed by many E. coli pathogens, but how it impacts hybrid strains has not been ascertained. Here, using zebrafish larvae as surrogate hosts to model both gut colonization and extraintestinal infections, we found that ECP can act as a multi-niche colonization and virulence factor for BA1250. Furthermore, our results indicate that ECP-related changes in activation of envelope stress response pathways may alter the fitness of BA1250. Using an in silico approach, we also delineated the broader repertoire of adhesins that are encoded by BA1250, and provide evidence that the expression of at least a few of these varies in the absence of functional ECP.

KEYWORDS:

• Escherichia coli
• hybrid
• zebrafish
• adhesin
• colonization, ECP
• ExPEC
• EPEC
• envelope stress

Introduction

The DNA sequencing era has improved our understanding of the factors and mechanisms involved in the diversification of Escherichia coli. Via the horizontal transfer of genes encoding virulence factors, E. coli became a highly diverse and versatile pathogen responsible for a variety of intestinal and extraintestinal diseases in humans and animals. Strains that are responsible for intestinal infections are known as diarrheagenic E. coli (DEC), a group that is further subdivided into pathotypes according to specific combinations of virulence traits and pathogenic mechanisms^1–3. Among DEC is an emerging pathotype known as atypical enteropathogenic E. coli (aEPEC)^4,5, which has the ability to cause severe diarrhea in both children and adults^6–8. Distinct from DEC are Extraintestinal pathogenic E. coli (ExPEC) strains, which are responsible for about 80% of urinary tract infections (UTIs) that affect more than 150 million people annually^2,9,10. ExPEC are also a leading cause of bloodstream infections that can lead to more severe, life-threatening conditions like sepsis and neonatal meningitis^11,12. The subset of ExPEC strains that are associated with meningitis in newborns are known as Neonatal Meningitis E. coli (NMEC), and are often lethal¹³. A primary reservoir for ExPEC strains is thought to be the lower intestinal tract, where ExPEC typically reside among the rest of the microbiota without eliciting any overt pathology^14–19.

Different E. coli pathotypes encode distinct sets of virulence factors, but horizontal gene transfer can result in the generation of strains that have a combination of virulence traits that are associated with both intestinal and extraintestinal pathotypes. These hybrid pathogenic strains often appear more virulent than their ancestral lineages¹, as exemplified by a hybrid (EHEC/EAEC) strain that caused a large diarrhea outbreak in Germany in 2011^20–22. In this study, we focus on a hybrid strain known as BA1250, which was isolated from a child with diarrhea in a case–control epidemiologic survey. BA1250 binds to host cells in vitro in a localized adherence-like pattern and was initially classified as atypical enteropathogenic E. coli (aEPEC)^23,24. aEPEC strains have a locus of enterocyte effacement (LEE) that encodes a type III secretion system, but they lack virulence determinants like bundle-forming pili that characterize typical EPEC isolates^7,25. Recently, BA1250 was reclassified as a hybrid aEPEC/ExPEC strain due to the presence of multiple genetic ExPEC-associated markers²⁶.

The transmission of aEPEC, ExPEC and other E. coli pathogens often occurs through fecal-oral contamination following contact with contaminated surfaces, food, or water^7,27. Upon entering the gastrointestinal tract, strains adhere to the mucosa of the large and small intestine where the main stages of pathogenesis have been described^5,28. Fimbrial and afimbrial adhesins are both essential for pathogenesis in the gut²⁹. Fimbriae (a.k.a. pili) are filamentous structures linked to the surface of the outer membrane that can mediate bacterial adhesion to eukaryotic cells^30–32. E. coli and other bacteria have a diversified repertoire of fimbrial adhesins with various known and hypothetical functions beyond adhesion, ranging from the capture of exogenous DNA to the evasion of antibiotics and the invasion of host cells^29,33–35.

In 2001, meningitis-associated and temperature-regulated (Mat) fimbriae were identified in NMEC as a potentially important temperature-controlled colonization factor³⁶. The same fimbriae were later identified in several DEC-associated pathotypes and renamed as E. coli common pilus (ECP)³⁷. ECP are assembled from a chaperone/usher-type pathway using six genes that comprise the ecp operon (see Figure 1a). The first gene, ecpR, encodes the regulator EcpR, which works with integration host factor (IHF) to promote transcription of the ecp operon³⁸. Next is ecpA, which encodes the major pilin subunit of ECP fimbriae, followed by ecpB. This gene and ecpE code for two chaperones, unusual for a chaperone/usher system which usually only utilize a single chaperone protein. The chaperones prevent misfolding and aggregation of pilin subunits within the periplasm prior to incorporation into growing pilus structures via usher complexes at the outer membrane³⁹. The ECP-associated usher protein is produced by the ecpC gene immediately downstream of ecpB, while the ECP-associated adhesin is encoded by ecpD just upstream of ecpE ³⁸.

Figure 1. Deletion of ecpA, but not removal of the entire ecp operon, impairs aEPEC/ExPEC colonization of the zebrafish gut. (a) Diagram of the ecp operon. (B and C) Graphs shows relative bacterial numbers recovered from zebrafish at the indicated times points following colonization of the gut by (b) BA1250, BA1250ΔecpA, or BA1250ΔecpRABCDE or by (c) BA1250/pRR48, BA1250ΔecpA/prr48, or BA1250ΔecpA/pEcpab. Horizontal lines indicate mean values, which are also denoted numerically below the graph. There was no statistical difference between the input titers for any of the strains. **, P < 0.01; ***, P < 0.001; **** P < 0.0001 by Mann Whitney U tests, relative to BA1250 or BA1250/pRR48 at the same time points; n = 15.

The importance of ECP has been examined using in vitro tests wherein deletion of ecpA greatly diminished the ability of DEC strains to bind HeLa cells^37,40,41. Other studies suggested that ECP, in addition to playing a role in host cell adherence by DEC, also contribute to biofilm formation by ExPEC strains. The role of ECP in biofilm formation appears to be dependent on surface hydrophobicity, as mutation of ecpA in NMEC strains led to reduced biofilm development on PVC and polystyrene surfaces, but not on glass⁴². Although ECP can clearly promote in vitro adhesion and biofilm formation by both DEC and ExPEC strains, there are no reports that assess ECP function in hybrid E. coli strains like BA1250, and the importance of ECP in vivo within varying host environments is likewise unclear.

Here, we developed zebrafish (Danio rerio) larvae as a relevant vertebrate model to examine the importance of ECP to the fitness and pathogenesis of the hybrid ExPEC/aEPEC hybrid strain BA1250 both within the intestinal tract and at extraintestinal sites. Due to their transparency, vertebrate physiology, mammalian-like host defenses, fecundity, and the availability of facile imaging and genetic tools, zebrafish have become a progressively popular model for analyzing bacteria-host interactions, immune responses, and microbial pathogenicity in vivo ^43–45. The use of zebrafish to examine bacterial infections, including those caused by DEC and ExPEC strains^43,46–52, has provided valuable insight into both host and microbial factors that promote disease processes. Results presented here reveal the importance of ECP as a multi-niche virulence and fitness determinant and shed new light on ECP regulation in relation to envelop stress responses and the expression of other adhesins.

Results

EcpA promotes colonization of the zebrafish intestinal tract

To examine how ECP impact the fitness and virulence potential of the aEPEC/ExPEC hybrid strain BA1250, we used the lambda Red recombinase system to create isogenic mutants lacking either the major pilin gene ecpA or the entire ecpRABCDE (Figure 1a)^53,54. We first tested the ability of BA1250 and its mutant derivatives to colonize the intestinal tract using zebrafish larvae at 4 d post-fertilization (dpf). At this stage of development, the larvae begin to swallow and have a functional digestive tract, allowing for waterborne infection by immersion in water containing bacteria of interest. Zebrafish larvae were co-incubated with the wild-type strain BA1250, BA1250ΔecpA, or BA1250ΔecpRABCDE in the fish water for 24 h and subsequently rinsed prior to continued incubations for another 1 to 3 d. In these assays, the wild-type strain stably colonized the larvae for the duration of the experiment, with a high bacterial load ranging between 1 × 10⁴ and 5 × 10⁵ Colony Forming Units (CFU) per fish (Figure 1b). Of note, in these assays, the vast majority of the bacteria are localized within the zebrafish gut (see Figure 2). BA1250ΔecpA was greatly impaired in its capacity to colonize zebrafish gut (Figure 1b). In the first 24 h post-inoculation (hpi), the recovered BA1250ΔecpA load was less than 10% of that observed with BA1250, and the ecpA mutant was nearly eliminated by 72 hpi. Surprisingly, BA1250ΔecpRABCDE colonized the zebrafish much like the wild-type strain at all time points examined, although titers of this mutant were more spread out. During the course of these gut colonization experiments, none of the fish infected with BA1250 or either mutant strain showed any overt signs of distress, and no deaths were observed. To validate that the colonization defect due to loss of ecpA was not attributable to any off-target mutations, BA1250ΔecpA was complemented with plasmid pEcpAB. This plasmid allows for the leaky expression of EcpA and its cognate chaperone EcpB, which was included to prevent misfolding and aggregation of the pilin subunits within the periplasm. The complemented strain was recovered at markedly higher levels from zebrafish at all time points than either the wild-type strain or BA1250ΔecpA carrying the empty vector control pRR48 (Figure 1c).

Figure 2. ECP impact the dynamic localization of BA1250 within the zebrafish gut. (a) Diagram showing the major regions of the zebrafish intestinal tract. (b) Fluorescent and merged light images showing location of BA1250, BA1250ΔecpA, and BA1250ΔecpRABCDE (all carrying pGEN-mCherry, red) at 24, 48, and 72 hpi of zebrafish, which were orally colonized at 96 hpf. The larvae are~3 mm long at this stage. White arrows indicate areas of bacterial colonization. (c) Graphs show levels of each bacterial strain, as determined by calculating Corrected Total Cell Fluorescence (CTCF), within the indicated regions of the gut over time. Bars indicate mean values ± SD. **, P < 0.002 by Mann Whitney U tests; n = 6. (d) CTFC values from each strain within the indicated intestinal regions plotted against time. Symbols in the line graphs denote mean values ± SD.

ECP alter the dynamic localization of BA1250 within the intestinal tract

Taking advantage of the transparent nature of zebrafish larvae and using fluorescently tagged bacterial strains carrying pGEN-mCherry, we assessed the localization of wild-type BA1250, BA1250ΔecpA, and BA1250ΔecpRABCDE over time within the three major regions of the zebrafish intestinal tract. These regions include the intestinal bulb, the mid-intestines, and the posterior intestines, as depicted in Figure 2a. Representative images showing localization of the wild-type and mutant strains at 24, 48, and 72 hpi are presented in Figure 2b. Graphs in Figure 2c show levels of fluorescent bacteria within the different intestinal regions quantified from multiple fish at each time point examined. The same data are graphed relative to time in Figure 2d, to provide a better sense of the dynamic spatiotemporal changes observed with each strain.

The majority of the wild-type strain BA1250 was localized within the intestinal bulb at 24 hpi, but shifted somewhat toward the posterior intestine by 72 hpi (Figure 2b-d). BA1250ΔecpA levels were also initially high within the bulb region, though notably less than either the wild-type strain or BA1250ΔecpRABCDE. At the same initial time point, BA1250ΔecpA levels were somewhat elevated within the posterior gut relative to both BA1250 and BA1250ΔecpRABCDE. By 48 and 72 hpi, BA1250ΔecpA was scarce throughout the intestinal tract. These observations reflect the titration data in Figure 1b and suggest that BA1250ΔecpA is rapidly cleared from the gut without establishing a niche where it can replenish its numbers. In contrast, BA1250ΔecpRABCDE was easily detected within the intestinal bulb region at all time points, surpassing the levels observed with BA1250 at 48 and 72 hpi (Figure 2b-c). Interestingly, unlike the wild-type strain, BA1250ΔecpRABCDE levels did not increase within the posterior region over time. The mid-intestinal region was sparsely colonized by all three strains, though BA1250ΔecpA and BA1250ΔecpRABCDE quantities were significantly different from BA1250 over time (Figure 2b-c). In total, these data indicate that while deletion of the entire ecp operon may not significantly affect overall bacterial titers (see Figure 1b), the lack of ECP expression can alter the dynamic localization of BA1250 within the gut.

ECP is an essential virulence factor during extraintestinal infections

Theoretically, hybrid aEPEC/ExPEC strains should be able to colonize intestinal and extraintestinal sites since they encode sets of fitness and virulence factor genes that are associated with both EPEC and ExPEC isolates. To test this idea, as well as the contributions made by ECP, we employed well-established zebrafish models of localized and systemic extraintestinal infections^46,48. To initiate localized infections, bacteria were microinjected into the pericardial cavity, where bacterial dissemination is restricted. Systemic infections were initiated by injection of bacteria directly into the bloodstream via the circulation valley. Following inoculation of the pericardial cavity, both wild-type BA1250 and BA1250ΔecpRABCDE killed nearly 70% of the larvae within 3 d (Figure 3a). In contrast, BA1250ΔecpA was completely avirulent in this model. Likewise, following inoculation into the bloodstream BA1250 and BA1250ΔecpRABCDE were similarly lethal (killing about 50% of the fish within 3 d), whereas BA1250ΔecpA was markedly less virulent (Figure 3b). These results demonstrate that the hybrid strain BA1250 has the capacity to cause serious disease outside of the intestinal tract, and highlights EcpA as a key regulator of fitness and virulence within extraintestinal niches.

Figure 3. EcpA promotes the virulence of BA1250 during both localized and systemic infections. At 48 hpf zebrafish were infected via inoculation of (a) the pericardial cavity or (b) the circulation valley with~1000 CFU/embryo of the indicated strains. Following inoculation, fish were scored for survival over the course of 72 h to generate Kaplan-Meier survival curves. P-values indicated were determined by Log-rank (Mantel-Cox) tests, versus BA1250. n = 68 fish.

ECP have no effect on motility, growth dynamics, or biofilm formation by BA1250

Differences observed between the ΔecpA and the ΔecpRABCDE strains in the pericardial and systemic infections reflect those observed in the gut colonization assays (Figures 1–3), and suggest that the two ecp mutants may differentially impact other fitness and virulence determinants. To begin to examine this possibility, we compared the wild-type and mutant strains in motility, growth and biofilm formation assays. Motility can have a significant impact on bacterial persistence within the host^55,56. Using in vitro assays with swim agar plates, we observed no differences in the motility of BA1250 or its derivatives BA1250ΔecpA and BA1250ΔecpRABCDE (Figure 4a). In these assays, Shigella flexneri served as a non-motile negative control. The wild-type and mutant BA1250 strains also did not vary in their ability to form biofilms, either on polystyrene or glass surfaces (Figure 4b). Surprisingly, in these assays all of the BA1250 strains lacked the ability to form appreciable biofilms in comparison with a reference biofilm-producing isolate, the enteroaggregative E. coli strain EAEC 042⁵⁷. All three BA1250 strains also grew similarly to one another in LB both at 37°C and in M9 minimal media at either 28.5°C or 37°C (Figure 4c). Importantly, none of the strains had any appreciable growth in fish water (E3) at 28.5°C. Together, these data indicate that deletion of ecpA alone or in combination with the other ecp genes does not alter the basic motility, growth, or biofilm formation characteristics of BA1250.

Figure 4. ECP have no effects on motility, growth dynamics or biofilm formation. (a) Graph shows spread of BA1250, BA1250ΔecpA, and BA1250ΔecpRABCDE on motility plates over time at 37°C. Shigella flexneri was used as a non-motile negative control. Mean values ± SD from two independent experiments done in triplicate are indicated. (b) Crystal violet biofilm assays were carried out in DMEM for 24 h or 48 h in polystyrene plates or on glass coverslips. EAEC 042, which forms robust biofilms, was used as positive control. Bars indicate mean values ± SD from three independent assays done in triplicate. (c) Graphs show growth curves of the indicated strains in LB, M9 or E3 at 37°C or 28.5°C, as indicated.

Deletion of ecpA versus ecpRABCDE elicits distinct types of envelope stress responses

Results from our zebrafish infection assays (Figures 1–3) indicate that BA1250ΔecpA and BA1250ΔecpRABCDE have divergent in vivo fitness and virulence phenotypes, even though both strains lack functional ECP. One explanation for this discrepancy may be that the deletion of ecpA causes dysregulated expression or aberrant assembly of other ECP components, which in turn could stress the bacteria and thereby alter their fitness within host environments. Specifically, we hypothesized that the deletion of ecpA might trigger more pronounced envelope stress responses than those occurring in either wild-type BA1250 or BA1250ΔecpRABCDE. To test this possibility, we employed three well-described expression reporter constructs in which the luciferase genes luxCDABE are under control of promoters for cpxP, spy, or rpoErseABC ⁵⁸. Expression of cpxP is driven by the CpxRA two-component envelope stress response system, which is activated by a variety of signals and envelope stresses, including the accumulation of some types of misfolded pilin subunits within the periplasm^59–62. The spy gene encodes a periplasmic chaperone that is also induced by the Cpx pathway, as well as by the BaeSR two component system which can respond to misfolded envelope proteins, membrane-damaging agents, and other signals^59,61,63. The rpoErseABC gene products constitute the ά^E envelope stress response pathway that can sense misfolded or mis-targeted outer membrane proteins or LPS within the periplasm^60,64. In addition to responding to envelope stress, each of these systems can also regulate the expression of a variety of adhesins and virulence factors within Gram-negative pathogens^60,64–66.

In growth assays done in DMEM, which induces ECP expression (see below), none of the envelope stress reporter constructs affected growth of BA1250, BA1250ΔecpA or BA1250ΔecpRABCDE over a 6-h time frame (Figure 5a-c). The kinetics and levels of cpxP:lux expression in BA1250ΔecpA mirrored that of the wild-type strain, while cpxP:lux was not notably induced at any time point in BA1250ΔecpRABCDE (Figure 5d). In contrast, levels spy:lux reporter were markedly higher in BA1250ΔecpRABCDE over time relative to both BA1250ΔecpA and BA1250 (Figure 5e). A similar trend was seen with the rpoErseABC:lux reporter (Figure 5f). These data indicate that BA1250ΔecpA and BA1250ΔecpRABCDE are experiencing distinct types of envelope stress, which could impact their overall fitness as well as patterns of virulence and adhesin gene expression.

Figure 5. Envelope stress responses activated in the ΔecpRABCDE mutant vary from those in the wild-type and ΔecpA mutant strains. The indicated lux reporters were transformed into BA1250, BA1250ΔecpA, and BA1250ΔecpRABCDE. Bacteria were grown in DMEM at 37°C, and (a-c) OD₆₀₀ and (d-f) reporter gene expression (luminescence) were measured over time. Luminescence levels were adjusted to OD₆₀₀. Data points indicate mean values ± SD of two independent assays, each with three replicates.

Does deletion of ecp genes alter the expression of other adhesins?

One explanation for the wild type-like colonization and virulence phenotypes displayed by BA1250ΔecpRABCDE is that the complete absence of the ecp operon may induce the compensatory expression of other adhesin(s). To address this question, we examined how deletion of ecpA or the entire ecp operon affects expression of five aEPEC-associated adhesive structures: type 1 pili (T1P), Hcp, curli, YcbQ, and intimin. Experiments were performed in DMEM at 28.5°C and 37°C (Figure 6 a and b, respectively) and in M9 minimal medium at 37°C (Figure 6c). The latter reflects conditions used to grow bacteria prior to use in the zebrafish infection models. Using Enzyme-linked immunosorbent assays (ELISAs), we observed a slight but significant difference in the expression of curli by BA1250ΔecpRABCDE in DMEM at 28.5°C, relative to the wild-type strain (Figure 6a). Significant, though modest, changes were also observed with YcbQ expression by BA1250ΔecpRABCDE in DMEM at 37°C (Figure 6b). Deletion of ecpA led to upregulation of T1P, Hcp, curli, and YcbQ in DMEM at 28.5°C and 37°C, with Hcp showing the greatest change (Figure 6 a-b). Furthermore, Hcp levels at 37°C in M9 were significantly elevated in BA1250ΔecpA relative to the wild-type strain, while T1P and YcbQ levels were notably lower (Figure 6c). Of note, ECP were not induced in M9 media at 37°C, but expression levels were greatly elevated in DMEM at both temperatures. In total, these observations indicate that deletion of the entire ecp operon does not induce altered expression of T1P, Hcp, YcbQ, or intimin, suggesting that these factors do not compensate for lack of ECP in our zebrafish models. On the other hand, upregulation of multiple adhesins, including Hcp, by BA1250ΔecpA may be detrimental to the overall fitness and virulence of this mutant within the zebrafish host.

Figure 6. Differential adhesin production associated with ECP expression. Expression levels of the indicated adhesins by BA1250, BA1250ΔecpA, and BA1250ΔecpRABCDEwere quantified by ELISA following growth in (a) DMEM at 28.5°C, (b) DMEM at 37°C, or (c) M9 at 37°C. Bars indicate mean values ± SD from three independent experiments with duplicate samples. **, P < 0.002 and *, P < 0.03, relative to BA1250 as determined by Mann Whitney Utests.

The aEPEC/ExPEC hybrid strain with significant adhesin machinery

To assess the spectrum of additional adhesins that might be differentially expressed in the absence of ECP components, an in silico analysis was performed with SPAAN software⁶⁷. This software applies machine learning to predict which amino acid sequence belongs to an adhesin protein and applies a value (Pad) for reliability, with values >0.51 considered trustable results and those that score >0.70 being ideal. SPAAN identified 303 putative adhesin proteins in BA1250 with Pad values >0.70. Among these, 91 are previously undescribed adhesins denoted as “Hypothetical proteins” and 31 are predicted to be fimbrial adhesins (Table 1). In addition to the adhesins noted and assayed above, other fimbrial adhesins detected with Pad scores >0.77 include Lpf, Sfa, Yfc, Yad, Pap, Yeh, K88, Ybg, CFA/I and Sfm (Table 2).

Table 1. Adhesin-like structures identified by SPAAN encoded by BA1250.

Characteristic	Pa*	Number of proteins
Proteins/Total adhesin	0.21–0.96	4.547
Adhesin	>0.51	1.045
Adhesin	>0.70	303
“Hypothetical protein”	>0.70	91
“Fimbrial”	>0.70	31

In silico analysis of BA1250 genome using SPAAN software to detect proteins with adhesin characteristics.

*P_ad>0,51 considered significant.

Table 2. Fimbrial adhesins identified at BA1250 genome by SPAAN.

Fimbrial Adhesin	Pad*
>DKEMMGOL_02627 putative major fimbrial subunit LpfA	0.96
>DKEMMGOL_03579 S-fimbrial protein subunit SfaH	0.94
>DKEMMGOL_03581 S-fimbrial protein subunit SfaG	0.86
>DKEMMGOL_00703 putative fimbrial-like protein YfcP	0.93
>DKEMMGOL_00704 putative fimbrial-like protein YfcQ	0.86
>DKEMMGOL_00708 putative fimbrial-like protein YfcV	0.78
>DKEMMGOL_00705 putative fimbrial-like protein YfcR	0.75
>DKEMMGOL_01394 putative fimbrial-like protein YadL	0.91
>DKEMMGOL_01395 putative fimbrial-like protein YadM	0.89
>DKEMMGOL_01398 putative fimbrial-like protein YadN	0.89
>DKEMMGOL_01392 putative fimbrial-like protein YadC	0.82
>DKEMMGOL_01393 putative fimbrial-like protein YadK	0.79
>DKEMMGOL_03313 Pap fimbrial major pilin protein	0.88
>DKEMMGOL_03307 Fimbrial protein PapE	0.79
>DKEMMGOL_03308 Fimbrial adapter PapK	0.76
>DKEMMGOL_03765 putative fimbrial-like protein YehA	0.87
>DKEMMGOL_04613 K88 fimbrial protein AC	0.87
>DKEMMGOL_01930 K88 fimbrial protein AD	0.77
>DKEMMGOL_01916 putative fimbrial-like protein YbgD	0.86
>DKEMMGOL_03306 Minor fimbrial protein PrsF	0.77
>DKEMMGOL_02112 CFA/I fimbrial subunit B	0.77
>DKEMMGOL_02633 putative fimbrial-like protein SfmH	0.77

*Fimbrial adhesins detected by SPAAN software, with Pad>0.70.

Discussion

Hybrid E. coli strains are pathogens that have genetic markers from more than one pathotype, which may include distinct adhesins, iron acquisition systems, and different classes of virulence factors⁶⁸. In pathogenic bacteria, fimbriae are often crucial adhesion factors that can mediate binding to target host cells, evasion of host defenses, host cell invasion, and biofilm formation^69,70. For these reasons, there is a lot of ongoing effort to better understand adhesin structures and their specific roles in colonization by different pathotypes. Despite the remarkable importance of fimbria in the infection process, most studies with hybrid E. coli pathotypes have been carried out only in vitro, employing cultured epithelial cells^71–73. Our understanding of how individual adhesins affect colonization processes and the virulence of hybrid strains in vivo within varying host environments is limited. For this reason, we set out to investigate the contributions of ECP to the fitness and virulence of the hybrid aEPEC/ExPEC strain BA1250 within the intestinal tract and at other extraintestinal sites of infection. Importantly, ECP are expressed by a highly heterogeneous group of other E. coli isolates in addition to BA1250, and so our findings are likely of relevance for a broad spectrum of E. coli pathogens^36,37.

Oral infection in the zebrafish model showed that deletion of ecpA greatly impairs BA1250 colonization of the gut, in line with results from in vitro cell culture-based assays^37,40. Taking a step further, we also showed that deletion of the entire ecp operon did not significantly impact bacterial colonization of the gut in terms of absolute bacterial numbers. This surprising result led us to consider more closely the regulation of ECP and potential effects on other bacterial systems, including envelope stress responses and the expression of other adhesins.

The ecp genes are transcribed as an operon from a promoter located 121 bp upstream of the initial ecpR codon, which is positively regulated by EcpR activation. The specific binding of EcpR leads to transcription of the rest of the genes that comprise ecp operon and the subsequent production of ECP³⁸. The observation that deletion of ecpA, but not removal of the entire ecp operon, impairs bacterial fitness within the zebrafish gut may be attributable to the regulator EcpR and its potential effects on other genes within the hybrid strain. The ability of EcpR to affect genes beyond the ecp operon is exemplified by a study in the neonatal meningitis strain IHE 3034 in which EcpR was shown to repress the expression of genes involved in flagella production, resulting in impaired motility⁷⁴. It is feasible that the absence of EcpR in BA1250ΔecpRABCDE modifies the expression of other factors (e.g. adhesins) that compensate for a lack of ECP, whereas dysregulation of EcpR within BA1250ΔecpA might stimulate the expression of other genes that are detrimental to BA1250 persistence within the gut. It is also possible that the marked changes in envelope stress responses observed with BA1250ΔecpRABCDE (see Figure 5) could contribute to modifications in gene expression patterns that compensate for loss of functional ECP, independent of EcpR. Discerning if the presence or absence of EcpR, or if differential activation of envelope stress responses, account for the fitness and virulence differences observed with BA1250ΔecpRABCDE and BA1250ΔecpA requires additional research.

Wang et al. performed a microarray-based analysis of the zebrafish gut and identified the presence of distinct regions with molecular characteristics of the mammalian small and large intestines⁷⁵. In our study, we noted that deletion of ecpA impaired BA1250 colonization throughout all sections of the zebrafish intestinal tract. Conversely, BA1250 and BA1250ΔecpRABCDE both initially colonized the bulb portion of the intestines at high levels. However, over time the wild-type strain shifted more toward the posterior regions of the gut, while BA1250ΔecpRABCDE remained mostly in the bulb (see Figure 2). These data indicate that the dynamic localization of BA1250ΔecpRABCDE within the gut may be influenced by differential expression of adhesins that vary from those produced by the wild-type strain or BA1250ΔecpA. Different sets of adhesins can alter the pattern of bacterial interactions with host cells and receptors⁷⁶. The types of host receptors and cells (e.g. enterocytes, mucin-producing goblet cells, and enteroendocrine cells) vary within the different sections of the gastrointestinal tract, and can thereby influence bacterial tropism based on the expression patterns of specific adhesins⁷⁷.

In an attempt to identify if there is altered expression of specific adhesins in BA1250ΔecpRABCDE, ELISAs were used to quantify levels of T1P, Hcp, Curli, Ycb, and Intimin in the wild-type and mutant strains. In these assays, deletion of the ecp operon had nearly no effect on expression of any of the tested adhesins. However, BA1250 encodes an arsenal of other adhesins that could conceivably be influenced directly or indirectly by deletion of the ecp genes²⁶. This point is driven home by our in silico analysis with SPAAN software, which indicated the presence of at least 303 genes within the BA1250 genome that encode proteins with adhesin-like characteristics. Among these are 31 fimbrial adhesins, including Lpf, Pap, CFA/I and S pilus systems that were previously shown to promote bacterial colonization of host tissues^78–82. Interestingly, 91 of the putative adhesins that were identified in BA1250 are uncharacterized hypothetical proteins and require further work to assess functionality.

Though expression of the specific adhesins that we assayed by ELISA were not altered in BA1250ΔecpRABCDE, production of T1P, Curli, Ycb, and Hcp were all significantly elevated in BA1250ΔecpA in at least one of the tested conditions. Upregulation of these adhesins may be a mechanism to help compensate for lack of functional ECP. However, if this is the case, the upregulation of these adhesins is not sufficient to rescue the colonization defects observed with BA1250ΔecpA in the zebrafish host. Alternatively, the dysregulated expression of adhesive organelles like Hcp may contribute to the defects observed with BA1250ΔecpA. Specifically, the aberrant overproduction of adhesins could potentially limit the ability of the hybrid strain to disseminate within the host, make the bacteria more susceptible to phage or host defenses, or reduce overall fitness due to the high energy costs associated with the biosynthesis of adhesive organelles⁸³. In considering these possibilities, it is noteworthy that deletion of ecpA either alone or with the entire ecp operon did not alter bacterial motility, growth kinetics, or biofilm development in our in vitro assays.

Bacterial pathogens can often survive in varying and volatile environments by sensing and adapting to changing environmental cues and altering their gene expression patterns⁸⁴. ECP expression has been detected in bacteria both within the intestines and at extraintestinal sites such as the urinary tract^36,37,40, suggesting that these adhesive organelles function within diverse niches. To assess the capacity of ECP expression to promote the survival and virulence of the hybrid strain BA1250 within distinct extraintestinal niches, wild-type and ecp mutant bacteria were inoculated into the pericardial cavity or the bloodstream of zebrafish larvae to initiate localized or systemic infections, respectively. In these models, virulence often correlates with the ability of E. coli pathogens to persist and replicate within the zebrafish host^46,48. Wild-type BA1250 and BA1250ΔecpRABCDE were similarly lethal in both the localized and systemic infection models, while inoculation of BA1250ΔecpA into either niche caused little, if any, host death (see Figure 3). These results mirror those from our zebrafish gut colonization assays, implicating ECP as important mediators of fitness and virulence within distinct host environments.

In total, this work demonstrates that hybrid aEPEC/ExPEC strains like BA1250 are versatile pathogens capable of effectively colonizing both intestinal and extraintestinal niches. Furthermore, our results highlight the utility of zebrafish as surrogate hosts that can be used to identify and characterize fitness and virulence traits associated with hybrid strains. Though it is clear from our data that ECP are critical colonization and virulence determinants for the aEPEC/ExPEC hybrid strain BA1250, the possibility that ECP are coordinately regulated with other adhesins like Hcp underscores the complexity of the bacterial adhesion process. Discerning the direct and indirect effects of ECP expression on bacterial fitness and virulence, including links with other adhesins and envelope stress responses, will require additional work.

Material and methods

Ethics statement

Zebrafish were bred at the Centralized Zebrafish Animal Resource (CZAR) at the University of Utah, and used in accordance with standardized protocols that were approved by the University of Utah Institutional Animal Care and Use Committee (IACUC).

Bacterial strains and plasmids

The strains and plasmids used in this study are listed in Table S1. BA1250 (eae+/EAF−/stx⁻²⁶/BFP−) has been described^23,24. Bacteria were cultivated from frozen stocks at 37°C for 16–18 h in static Luria-Bertani (LB) or M9 minimal medium (6 g/l Na₂HPO₄, 3 g/L KH₂PO₄, 1 g/L NH₄Cl, 0.5 g/L NaCl, 1 mM MgSO₄, 0.1 mM CaCl₂, 0.1% glucose, 0.0025% nicotinic acid, 0.2% casein amino acids, and 16.5 mg/mL thiamine in H₂O), as indicated. Antibiotics were added to the growth medium as necessary: ampicillin (100 μg/mL), kanamycin (50 μg/mL), or chloramphenicol (25 μg/mL).

The deletion of ecpA and ecpRABCDE genes was performed using the λ Red recombinase system as previously described using plasmids and primers listed in Table S2^53,54. Specifically, the chloramphenicol resistance cassette (Cam^R) flanked by FRT sites in pKD3 was amplified by PCR using primers with 40 bp ends that are homologous to the 5’ and 3’ ends of ecpA (primers ecpAcat-1 and ecpAcat-2). The FRT-flanked kanamycin resistance cassette (Kan^R) was similarly amplified from pKD4 using primers with homology to the 5’ end of ecpR and the 3’ end of ecpE (primers ecpRkan-1 and ecpEkan-2). PCR products were electroporated into BA1250 cells carrying pKD46 or pKM208, which encode IPTG- and arabinose-inducible lambda Red recombinase, respectively^53,54. Knockouts were confirmed using PCR with primers listed in Table S2, and resistance cassettes were removed using the FLP flippase encoded by pCP20⁵³. Knockout mutants with the Cam^R or Kan^R cassettes present were used for all of the in vivo zebrafish assays, while mutants with the resistance cassettes flipped out were used for the luciferase reporter assays. Both sets of mutants were used for all other assays, with no discernable differences observed due to the presence or absence of either the Cam^R or Kan^R cassettes. Only results from in vitro assays done with mutant strains lacking the resistance cassettes are shown.

To generate pGEN-mCherry, mCherry sequences were amplified from pmCherry (Clontech) using primers that added a 5’ SnabI site and synthetic RBS, as well as a 3’ NotI site. The PCR product was ligated into pGEN-luxCDABE behind the constitutive Pem7 promoter, replacing the lux genes, using standard molecular biology approaches⁸⁵. The ecpAB genes were amplified from BA1250 using the ecpA-NdeI and ecpB-HindIII primers, which incorporated NdeI and HindIII sites that were used to insert ecpAB into pRR48 behind the IPTG-inducible tac promoter⁸⁶. Expression constructs were confirmed by Sanger sequencing.

Zebrafish infection assays

Wild-type *AB zebrafish embryos were collected from breeding colonies that were maintained on a 14 h/10 h light/dark photoperiod cycle. Embryos were grown at 28.5°C in E3 media (5 mM NaCl, 0.17 mM KCl, 0.4 mM CaCl₂, 0.16 mM MgSO₄) containing 0.000016% methylene blue as an anti-fungal agent. Following collection at 24 or 48 hpf, eggs were incubated for 5 min in sterile E3 media plus 0.0038% (v/v) of bleach solution (Sodium Hypochlorite and Chlorine; Sigma Aldrich) and then rinsed using two 5-minute washes with sterile double distilled water followed by two 5-minute washes in sterile E3 media. Embryos were then dechorionated using 1 mg/mL pronase (Roche, VWR) and maintained in sterile E3 media at 28.5°C prior to infection.

For intestinal colonization assays, bacteria were grown statically in M9 media for about 18 hours and then resuspended in E3 media. Zebrafish larvae at 96 hpf were transferred to fresh Petri dishes and immersed in E3 media containing bacteria (~1 × 10⁹ CFU/mL) for 24 h. The fish were then rinsed with sterile E3 and placed in fresh E3 media every 24 h over the course of 3 d. Sets of larvae were euthanized at 24, 48, and 72 hpi and subsequently homogenized in PBS using a Bullet Blender tissue homogenizer (Next Advance, USA). Serial dilutions of the homogenates were plated on MacConkey agar and incubated at 37°C for ~18 hours to determine bacterial numbers per larvae. Complementation assays were performed similarly with the exception that upon infection and subsequent E3 changes, the E3 contained ampicillin to maintain the plasmids and 1 mM IPTG to induce expression of EcpAB.

For localized or systemic extraintestinal infections, wild-type or mutant bacteria (~1000 CFU in 1 nL of PBS per larvae) were microinjected into either the P.C. or bloodstream, as previously described^46,48. The fish were then monitored over the course of 3 d for signs of disease and viability. The absence of a heartbeat and blood flow were used the primary indicators of death.

Zebrafish imaging

Zebrafish larvae at 96 hpf were immersed in E3 media containing ~1 × 10⁹ CFU/mL of BA1250, BA1250ΔecpA, or BA1250ΔecpRABCDE that were transformed with pGEN-mCherry. At 24, 48, and 72 hpi, larvae were carefully washed with PBS to remove any external bacteria, and then immobilized in 3% methylcellulose placed on 3% agar within a Petri dish. Animals were visualized using a fluorescent Olympus SZX10 microscope equipped with an Olympus DP72 camera. Fluorescent signals from captured images were quantified using Fiji/ImageJ 2 2.3.0/1.53f⁸⁷. The corrected total cell fluorescence (CTCF) was calculated using the formula: CFCT = integrated density − (selected area × mean fluorescence of background readings)⁸⁸. For each image, three selected and background areas were used to normalize against autofluorescence.

Bacterial motility

Bacterial strains were cultivated statically in LB at 37°C for 18 h. After this period, 10 µL aliquots of each bacterial culture were applied to the center of Petri dishes containing LB semi-solid medium (0.3% agar). These plates were incubated at 37°C and motility was assessed at 6, 12, 16, 18, and 20 h time points by measuring the spread of bacteria around the initial inoculation site. Shigella flexneri, which is immotile, was used as a negative control in these assays. The assays were performed in triplicate with two independent repetitions.

Growth curves

Bacteria were grown statically from frozen stocks in 3 mL of LB for 18 h at either at 28.5°C or 37°C. Afterward, bacteria were diluted 1:100 into LB, M9, or E3 media and 200-μL aliquots of each sample were transferred to 100-well honeycomb plates (Thermo Scientific, UK). Growth curves were obtained using a Bioscreen C instrument (Growth Curves USA) with gentle shaking to acquire OD₆₀₀ readings every 30 min for 24 h. These experiments were performed with quadruplicate samples.

Crystal violet biofilm assays

Adhesion and biofilm formation assays on polystyrene and glass surfaces were performed as described by Sheikh et al.⁸⁹, with some modifications. Strains were grown statically in TSB medium for 18 h at 37°C and then diluted 1:100 into modified DMEM medium without phenol (CultiLab). Aliquots (200 μL) were distributed into 96-well polystyrene cell culture plates (TPP) or into 24-well cell culture plates containing glass coverslips. The plates were then incubated statically at 37°C for 24 or 48 h and each well was subsequently washed with PBS prior to fixation in 75% ethanol for 10 min. Wells were washed an additional 3 times with PBS, stained with a 0.5% crystal violet solution for 5 min, and rinsed 3 more times with PBS. Crystal violet that was retained within any biofilm communities was solubilized using 95% ethanol and quantified by measuring absorbance (595 nm) using a Multiskan EX plate reader (Thermo Fisher Scientific, USA). These assays were performed in triplicate with three independent replicates. EAEC 042, which forms robust biofilms, was used as positive control.

Envelope stress assays

The cpxP-, spy-, and rpoErseABC-luxCDABE reporter constructs (pJW1, pNLP15, and pNLP19, respectively)⁵⁸ were transformed into wild-type and mutant BA1250 strains. These recombinant strains were grown statically from frozen stocks for about 18 hours in M9 minimal media containing kanamycin. Cultures were washed twice with PBS (Hyclone) and then diluted 1:100 in DMEM without phenol red (Life Technologies). 100 μL-aliquots of each strain were added individually to wells of a sterile, white 96-well plate (Thermo Scientific) and luminescence readings were acquired every 30 min for 9 h at 37°C using Gen5 software with a BioTek Synergy H1 plate reader. Before each measurement the plate was shaken for 1 second. For luminescence emission, the signal was read over 10 seconds with a gain of 135 and height of 1 mm, as described previously¹⁷. To account for any differences in growth, OD₆₀₀ was read at the same time. This assay was performed using two biological replicates, each repeated three times.

Fimbrial adhesin detection

BA1250, BA1250ΔecpA, and BA1250ΔecpRABCDE were grown statically in LB at 37°C for 16–18 h, sub-cultured 1:100 into DMEM medium or M9 minimal medium, and then incubated for an additional 18 h at 28.5°C or 37°C. Next, the bacteria were pelleted by centrifugation at 5,000 × g for 10 min and resuspended in 0.05 M Carbonate Bicarbonate buffer (Ca/Bi Buffer, pH 9.6) to obtain an optical density (OD₆₀₀) of 1.0. Alternatively, for assays designed to detect intimin, bacterial pellets were first resuspended in a 4% Triton X – 100 solution for 5 min, pelleted, and then resuspended in Ca/Bi buffer. Bacterial suspensions were used to coat MaxiSorp plates (NUNC – Thermo Fisher Scientific) at room temperature for 18 h. The plates were washed three times with PBS with 0.05% Tween-20 (PBS-T), blocked with a 1% solution of bovine serum albumin (BSA) in PBS for 30 min at room temperature, and then and incubated for 1 h at room temperature with the following primary antibodies: rabbit anti-ECP serum (1:3,000)³⁷, rabbit anti-T1P serum (1:5000) (kindly donated by Dr Jorge Girón), rabbit anti-HCP serum (1:5,000)⁹⁰, rabbit anti-Curli serum (1:5000)⁹¹, rabbit anti-YcbQ serum (1:5,000)⁹², or anti-intimin IgG enriched fraction (1:2,000)⁹³ in blocking solution. Plates were again washed with PBS-T and incubated for 1 h at room temperature with peroxidase-conjugated goat anti-rabbit IgG antibodies (1:5,000) in blocking solution. Finally, the plates were washed with PBS-T, developed using σ-nitrophenyl-β-D-galactopyranoside (OPD; Sigma-Aldrich), and OD₄₉₂ values were acquired using a Multiskan EX plate reader. Background signals were determined using wells that were treated with Ca/Bi buffer alone (without bacteria), and subtracted from corresponding samples which contained bacteria. All strains and adhesins were tested in duplicate in 3 independent assays.

In silico analysis of adhesin genes

To search for potential adhesive structure genes, coding DNA sequences (CDS) within the BA1250 genome were analyzed using the SPAAN software⁶⁷. This software uses artificial intelligence via the application of a specific type of machine learning algorithm referred to as an artificial neural network. The neural network is trained to recognize adhesive structures using data from known adhesins obtained from the NCBI website (https://www.ncbi.nlm.nih.gov/). To make predictions, the software considers five key features of the amino acids sequences encoded by BA1250: amino acid frequency, amino acid repeat frequency, dipeptide frequencies, electrical charge composition, and hydrophobicity. Cumulatively, these characteristics comprise an input layer that is processed by the neural network and transferred to other layers to eventually provide predictions of the probability that a specific CDS encodes an adhesin.

Statistical analyses

Kolmogorov–Smirnov tests were performed for the determination of non-parametric datasets, Mann-Whitney U tests for colonization and adhesin production assays, and Log-rank (Mantel-Cox) tests for the zebrafish survival assays. P values of<0.05, as determined by Prism v5.0 (GraphPad Software), were considered significant.

Author contributions

Conceived and designed the experiments: DDM ACR MAM RMFP. Performed the experiments: DDM ACR FFS. Analyzed the data: DDM ACR FFS MAM RMFP. Contributed reagents/materials/analysis tools: MAM RMFP. Wrote the paper: DDM ACR MAM RMFP.

Acknowledgments

We thank Fundação Butantan for financial support and personnel from Centralized Zebrafish Animal Resource (CZAR) at the University of Utah, Salt Lake City, Utah, United States for their advice and help with maintaining the zebrafish breeders. We are also grateful to Dr. Jorge Girón for the kind donation of adhesin antibodies, Dr. T. Raivio luxCDABE reporter constructs, and Dr. Travis Wiles for help with creating pGEN-mCherry.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Some data underlying this article are available in the GenBank database under the accession number JADPBX000000000, BioProject and SRA data PRJNA678986, at https://www.ncbi.nlm.nih.gov/genbank/, and https://doi.org/10.11606/T.42.2020.tde-04012022–135948.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19490976.2023.2190308.

Additional information

Funding

This work was supported in part by NIH grants R21 AI135918 and R01 GM134331 to M.A.M and NIH training grant from T32 AI055434 to A.C.R; Brazilian Federal Agency for Support and Evaluation of Graduate Education – CAPES under Grant PDSE - 88881.186835/2018-01, and CNPq under Grant 140088/2016-4 to D.D.M; and São Paulo Research Foundation under grant FAPESP 2017/25406-0 to R.M.F.P.