Abstract
Enterohemorrhagic Escherichia coli (EHEC) colonize intestinal epithelium by generating characteristic attaching and effacing (AE) lesions. They are lysogenized by prophage that encode Shiga toxin 2 (Stx2), which is responsible for severe clinical manifestations. As a lysogen, prophage genes leading to lytic growth and stx2 expression are repressed, whereas induction of the bacterial SOS response in response to DNA damage leads to lytic phage growth and Stx2 production both in vitro and in germ-free or streptomycin-treated mice.
Some commensal bacteria diminish prophage induction and concomitant Stx2 production in vitro, whereas it has been proposed that phage-susceptible commensals may amplify Stx2 production by facilitating successive cycles of infection in vivo. We tested the role of phage induction in both Stx production and lethal disease in microbiome-replete mice, using our mouse model encompassing the murine pathogen Citrobacterrodentium lysogenized with the Stx2-encoding phage Φstx2dact. This strain generates EHEC-like AE lesions on the murine intestine and causes lethal Stx-mediated disease. We found that lethal mouse infection did not require that Φstx2dact infect or lysogenize commensal bacteria. In addition, we detected circularized phage genomes, potentially in the early stage of replication, in feces of infected mice, confirming that prophage induction occurs during infection of microbiota-replete mice. Further, C. rodentium (Φstx2dact) mutants that do not respond to DNA damage or express stx produced neither high levels of Stx2 in vitro or lethal infection in vivo, confirming that SOS induction and concomitant expression of phage-encoded stx genes are required for disease. In contrast, C. rodentium (Φstx2dact) mutants incapable of prophage genome excision or of packaging phage genomes retained the ability to produce Stx in vitro, as well as to cause lethal disease in mice. Thus, in a microbiome-replete EHEC infection model, lytic induction of Stx-encoding prophage is essential for lethal disease, but actual phage production is not.
Author summary Enterohemorrhagic Escherichia coli (EHEC), a food-borne pathogen that produces Shiga toxin, is associated with serious disease outbreaks worldwide, including over 390 food poisoning outbreaks in the U.S. in the last two decades. Humans acquire EHEC by ingesting contaminated food or water, or through contact with animals or their environment. Infection and toxin production may result in localized hemorrhagic colitis, but may progress to life-threatening systemic hemolytic uremic syndrome (HUS), the leading cause of kidney failure in children. Treatment for EHEC or HUS remains elusive, as antibiotics have been shown to exacerbate disease.
Shiga toxin genes reside on a dormant bacterial virus present in the EHEC genome, but are expressed when the virus is induced to leave its dormant state and begin to replicate. Extensive virus replication has been thought necessary to produce sufficient toxin to cause disease.
Using viral and bacterial mutants in our EHEC disease mouse model, we showed that whereas an inducing signal needed to begin viral replication was essential for lethal disease, virus production was not: sufficient Shiga toxin was produced to cause lethal mouse disease, even without viral replication. Future analyses of EHEC-infected human samples will determine whether this same phenomenon applies, potentially directing intervention strategies.
Introduction
Shiga toxin-producing Escherichia coli (STEC) is a food-borne zoonotic agent associated with worldwide disease outbreaks that pose a serious public health concern. Enterohemorrhagic Escherichia coli (EHEC), a subset of STEC harboring specific virulence factors that promote a specific mode of colonization of the intestinal epithelium (see below), is acquired by humans by ingestion of contaminated food or water, or through contact with animals or their environment. EHEC serotype O157:H7 is a major source of E. coli food poisoning in the United States, accounting for more than 390 outbreaks in the last two decades[1–5]. EHEC infection usually presents as localized hemorrhagic colitis, and may progress to the life-threatening systemic hemolytic uremic syndrome (HUS), characterized by the triad of hemolytic anemia, thrombocytopenia, and renal failure [5, 6]. HUS is the leading cause of renal failure in children [7].
EHEC, along with enteropathogenic E. coli and Citrobacter rodentium belong to the family of bacteria known as attaching and effacing (AE) pathogens that are capable of forming pedestal-like structures beneath bound bacteria by triggering localized actin assembly [8–10]. While this ability of EHEC leads to colonization of the large intestine, production of prophage-encoded Shiga toxin (Stx) promotes intestinal damage resulting in hemorrhagic colitis [11–17]. Shiga toxin may further translocate across the colonic epithelium into the bloodstream, leading to systemic disease. Distal tissue sites, including the kidney, express high levels of the Shiga toxin-binding globotriosylceramide (Gb3) receptor, potentially leading to HUS [14, 15, 18–21].
Genes encoding EHEC Shiga toxin are typically encoded in the late gene transcription region of integrated lambdoid prophages [22, 23] and their expression is thus predicted to be temporally controlled by phage regulons [24–27]. Early studies showed that high levels of Stx production and release from the bacterium in vitro required prophage induction, i.e., the mechanism by which quiescent prophages of lysogenic bacteria are induced to replicate intracellularly and released as phage particles by host cell lysis [27, 28]. Lambdoid phage inducers are most commonly agents that damage DNA or interfere with DNA synthesis, such as ultraviolet light or mitomycin C. These inducing stimuli trigger activation of the bacterial RecA protein, ultimately leading to the cleavage of the prophage major repressor protein, CI, allowing expression of phage early and middle genes. Late gene transcription, which requires the Q antiterminator, results in the expression of many virion structural genes and of endolytic functions S and R, which lyse the bacterium and release progeny phage [29]. Other signaling pathways involving quorum sensing or stress response have also been implicated in lysogenic induction [30, 31].
Unfortunately, antibiotics commonly used to treat diarrheal diseases in children and adults are known to induce the SOS response‥ Trimethoprim-sulfamethoxazole and ciprofloxacin have been shown to enhance Stx production in vitro [32–34], and antibiotic treatment of EHEC-infected individuals is associated with an increased risk of HUS [35]. Hence, antibiotics are contraindicated for EHEC infection and current treatment is limited to supportive measures [36].
A more detailed understanding of the role of prophage induction and Stx production and disease has been pursued in animal models of EHEC infection. Although some strains of conventional mice can be transiently colonized with EHEC, colonization is not robust and typically diminishes over the course of a week [13, 37], necessitating use of streptomycin-treated [16] or germ-free mice [38, 39] to investigate disease manifestations that require efficient, longer term intestinal colonization. In streptomycin-treated mice colonized with EHEC, administration of ciprofloxacin, a known SOS inducer, induces the Stx prophage lytic cycle, leading to increased Stx production in mouse intestines and to Stx-mediated lethality [40]. Conversely, an EHEC strain encoding a mutant CI repressor incapable of inactivation by the SOS response was also incapable of causing disease in germ-free mice [41].
A potential limitation of the antibiotic-treated or germ-free mouse infection models is the disruption or absence, respectively, of microbiota, with concomitant alterations in immune and physiological function [42]. For example, a laboratory-adapted E. coli strain that lacks the colonization factors of commensal or pathogenic E. coli is capable of stably colonizing streptomycin-treated mice [43], and, when overproducing Stx2, is capable of causing lethal infection in antibiotic-treated mice [17]. Further, as up to 10% of human gut commensal E. coli were found to be susceptible to lysogenic infection by Stx phages in vitro [44], and it has been postulated that commensals may play an amplifying role in EHEC disease by fostering successive rounds of lytic phage growth [44–47]. Finally, gut microbiota may also directly influence expression of stx genes. For example, whereas a genetic sensor of phage induction suggests that the luminal environment of the germ-free mouse intestine harbors a prophage-inducing stimulus [41], several commensal bacteria have been shown to inhibit prophage induction and/or Stx production in vitro [48–50]. Alternatively, colicinogenic bacteria produce DNAse colicins that may trigger the SOS response, increasing Stx production [51].
Our laboratory previously developed a murine model for EHEC using the murine AE pathogen C. rodentium [52, 53], which efficiently colonizes conventionally raised mice and allows the study of infection in mice with intact microbiota. The infecting C. rodentium is lysogenized with E. coli Stx2-producing phage Φ1720a-02 [52, 54] encoding Stx variant Stx2dact, which is particularly potent in mice [55, 56]. Infection of C57BL/6 mice with C. rodentium (Φ1720a-02), (herein referred to as C. rodentium (Φstx2dact)), produces many of the features of human EHEC infection, including colitis, renal damage, weight loss, and potential lethality, in an Stx2dact-dependent manner [52].
We have been interested in phage, bacterial, and host factors that lead to lethal EHEC infection. In the current study, we found that C. rodentium (stx2dact) strains lacking RecA, which is required for induction of an SOS response, or phage Q protein, which is required for efficient transcription of the late phage genes, did not produce high levels of Stx in vitro or cause lethal disease in mice. In contrast, mutants defective in prophage excision, phage assembly, or phage-induced bacterial lysis retained the ability to both produce Stx upon prophage induction in vitro and to cause lethal disease. Excised phage genomes, potentially undergoing DNA replication leading to phage production or representing packaged phage, were detected, albeit at low levels, in fecal samples of mice infected with wild type C. rodentium (φStx), but not in mice infected with excision-defective C. rodentium (φStx). Thus, in a microbiome-replete EHEC infection model, lytic induction of Stx-encoding prophage, but not actual production of viable phage particles, is essential for Stx production and lethal disease.
Results
Gene Map and Features of Φstx2dact prophage
Lambdoid phage Φ1720a-02 was originally isolated from EC1720a-02, a STEC strain found in packaged ground beef [54]. To create the novel C. rodentium-mediated mouse model of EHEC infection, C. rodentium DBS100 (also known as C. rodentium strain ICC 168 (GenBank accession number NC_013716.1)), was lysogenized with phage Φ1720a-02 marked with a chloramphenicol (cam)-resistance cassette inserted into the phage Rz gene, creating lysogen DBS770 [52, 53]. A second lysogen, DBS771, encodes the chloramphenicol-marked prophage with an additional kanamycin (kan)-resistance cassette inserted into the prophage stx2A gene. Thus, in contrast to strain DBS770, DBS771 is unable to produce Shiga toxin or mediate lethal infection in mice. For simplicity, and for clarity with regard to phage mutations, phage Φ1720a-02, and strains DBS770 and DBS77′ will herein be referred to as Φstx2dact, C. rodentium (Φstx2dact), and C. rodentium (ΦΔstx2dact::kanR), respectively (Table 1).
To identify phage genes critical for lethal mouse infection, we sought to inactivate specific prophage genes and then assess their resulting phenotypes in the C. rodentium mouse model. As a first step, we sequenced the parental strain DBS100 and the genomes of C. rodentium (Φstx2dact) and C. rodentium (Φstx2dact::kanR), revealing that the three genomes were identical except for the prophages present in C. rodentium (Φstx2dact) and C. rodentium (Φstx2dact::kanR) (data not shown). We utilized these sequences to annotate the entire Φstx2dactprophage (GenBank accession number KF030445.1) present in strains C. rodentium (Φstx2dact) and C. rodentium (Φstx2dact::KanR), as diagrammed in Fig. S1. As is typical of Stx phages, the prophage sequence revealed a lambdoid phage with a mosaic gene organization, but nevertheless syntenic to varying degrees with other lambdoid phages [65].
Of note, as is the case for another lambdoid phage (E. coli phage mEP460, GenBank accession number JQ182728), the orientation of the Φstx2dact central regulatory region encoding CI repressor, other regulators such as N, Q and Cro, and the lytic promoters PL and PR, is inverted with respect to the canonical map of phage lambda [66]. Although strains C. rodentium (Φstx2dact) and C. rodentium (Φstx2dact::kanR) were lysogenized independently, in both strains the prophage was integrated at the same location, i.e. 100 bp into the coding sequence of dusA, which encodes tRNA-dihydroxyuridine synthase A. A recent study [67] revealed that known integrase genes, at least half of which belong to prophages, were found adjacent to the host dusA gene in over 200 bacterial species. Furthermore, a 21 base pair motif found at the prophage-host DNA junctions in many bacteria was present at the prophage junctions, attL and attR, of C. rodentium (Φstx2dact) and C. rodentium (ΦΔstx2dact::kanR), as well as at the presumed attB phage insertion site in the parental C. rodentium dusA gene (Fig. 1). A seven-base segment within this 21-base sequence is completely conserved between attL, attR, and attB and likely represents the ‘core’ sequence within which recombination occurs during integration or excision (Fig. 1, bolded sequence; [68]). Finally, although the Φstx2dact and ΦΔstx2dact::kanR prophages interrupt the dusA gene, they encode a 184 bp ORF (designated “ΦdusA’” in Fig. 1) that is in frame with the 3’ 937 nucleotides (positions 101 to 1038) of dusA that could serve to foster the production of a protein containing the C-terminal 312 amino acids of the canonical DusA.
A prior analysis of the host C. rodentium DBS100 genome sequence revealed the presence of 10 additional partial and intact prophages distributed around the genome [69]. Using NCBI BLAST nucleotide analysis, we found only 2 regions of homology between Φstx2dactand these prophages: one resident prophage (integrated at C. rodentium genome bp 2764517 – 2766051) encoded partial (70%) homology to the Φstx2dact cI repressor gene, and another resident prophage (integrated at C. rodentium genome bp 2097787-2098157) showed 79% homology to a Φstx2dact gene encoding a hypothetical protein (data not shown). No other significant homology between Φstx2dact and the resident prophages was detected.
Survey of prophage integration (attP) sites during murine infection reveals Φstx2dact prophage excision from C. rodentium (Φstx2dact), but no secondary lysogeny of commensal bacteria
In the course of EHEC infection of streptomycin-treated mice, Stx phage could be induced by antibiotic treatment to lysogenize other E coli strains in the intraluminal environment [40], (also see [70]). It has been postulated that successive cycles of infection of non-pathogenic commensal E. coli could amplify Stx production and exacerbate disease [38, 44, 45, 47]. We first addressed this question by testing whether lysogeny of commensal bacteria by phage Φstx2dact was detectable following oral C. rodentium (Φstx2dact) infection of mice. Mice orally gavaged with C. rodentium (Ostx2dact) normally exhibit weight loss and lethal disease ([52], not shown), typically succumbing to disease after day 7 post-infection. DNA was extracted from fecal samples of a group of five mice at days 1 and 6 post infection. The DNA samples were used as a template to generate a library of sequences encompassing the sequence downstream of attL (specifically, spanning the region from the phage int gene, through ΦdusA and into the adjacent host sequence; see Fig. 1). This strategy is a modification of that used for Tn-seq library analysis ([62], Materials and Methods).
Although we were unable to obtain detectable amplified DNA from fecal samples produced on day 1 post-infection, consistent with the low titer of C. rodentium (Φstx2dact) in the stool at this early time point, the day 6 post-infection sample yielded a DNA library, which was subjected to massively parallel sequencing to identify the origin of the host DNA into which the prophage was integrated. Of the total 17,868,095 sequences generated, all but 725,997 (4.06%) were of sufficient quality to analyze (Table 2; see Materials and Methods). Of the readable sequences, 99.56% showed homology to C. rodentium (Φstx2dact), i.e. included C. rodentium (Φstx2dact) attL and the adjacent C. rodentium dusA gene sequence, indicating prophage integration into the original C. rodentium strain. For the remaining 0.44% of sequences, the C. rodentium dusA sequences adjacent to the attL core sequence were found to be replaced by phage-specific sequences of attR at the other end of the prophage. Hence, these 0.44% of sequences encode attP, likely regenerated by recombination of attL and attR and probably reflecting excised circular phage genomes generated following induction of the C. rodentium (Φstx2dact) lysogen. Thus, our analysis indicates that C. rodentium (Φstx2dact) undergoes lytic induction in the murine host, consistent with previous findings of EHEC infection in streptomycin-treated mice. Furthermore, of the more than 17 million sequences analyzed, none showed integration of the Φstx2dact prophage into a different site in C. rodentium, or into a different bacterial host. Thus, lysogeny of commensal bacteria by Φstx2dact is not a common event in this model.
C. rodentium RecA and Φstx2dact proteins integrase, Q, endolysins, and portal protein are required for efficient phage production and release in vitro
Prophage induction of lambdoid phages is often initiated by DNA damage, in which SOS pathway activation leads to RecA-promoted autocleavage of CI repressor, followed by :ranscription of early genes from the from PL and PR promoters. Subsequent temporally programmed transcription of the prophage genome results in the production of delayed early (middle) proteins such as Int (integrase), essential for prophage integration and excision, and antiterminator protein Q. Production of Q in turn mediates the transcription of late genes, ncluding portal protein gene B, responsible for translocation of phage DNA into the virion protein capsid, and lysis genes S and R, encoding endolysins that disrupt the bacterial plasma membrane causing release of intact phage progeny (for a review, see Gottesman and Weisberg [66]). Late genes in EHEC phages also encompass stx.
To uncover the roles of specific phage and bacterial functions in EHEC disease, we used lambda red recombination (Materials and Methods) to construct C. rodentium (Φstx2dact) strains defective for prophage genes SR, int, B, or Q, and the host gene recA, which is well documented to be central to the SOS response and lytic induction. In addition, we inactivated three other genes that have been implicated as having more subtle roles in the lytic induction of Shiga toxin-encoding phage [30, 31]: rpoS, which controls the bacterial stress response, and qseC and qseF, which control quorum sensing pathways (Materials and Methods, Table 1). None of the mutants displayed a growth defect upon in vitro culture in rich broth (Fig. S2 and data not shown).
We then tested C. rodentium (Φstx2dact) and several of the mutant derivatives predicted to have dramatic effects on phage production for the ability to generate Φstx2dact following SOS induction. Pilot experiments revealed that Φstx2dact plaques were not detectable on any of numerous indicator strains (not shown), so phage production was measured by qPCR [71] using primers flanking the phage attP site (Table 3). As only excised phage have a reconstituted attP site [66], these primers only amplify product from unintegrated phage genomes.
In supernatants of LB cultures at mid-logarithmic growth phase, designated as t=0h, we detected 1.3⨯109 - 3.8⨯109 attP copies (phage genomes) per ml (Table 4 legend). The midlog cultures of wild type C. rodentium (Φstx2dact) contained only approximately 108 viable bacteria per ml, indicating that significant spontaneous prophage induction had occurred during the 2-4 hours of culture of this strain prior to entering mid-log growth. Four hours of further culture of C. rodentium (Φstx2dact) (t=4h) resulted in a 3.2-fold increase in the concentration of Φstx2dact in the supernatant compared to that at t=0h, consistent with continued spontaneous prophage induction (Table 4, “Relative attP production”, “- Mito C”). Prophage induction of the wild type lysogen with the SOS inducer mitomycin C led to a 234fold increase in relative attP production (Table 4, “+ Mito C”), a 73-fold increase above baseline levels. As predicted [72, 73], the generation of circular phage genomes required integrase; at all time points tested, attP copies were below the level of detection of 1⨯ 104/ml in uninduced or mitomycin C-induced cultures of the C. rodentium (Φstx2dactΔint) mutant, which lacks Int recombinase (Table 4).
Host and phage functions contributed to the amount of phage production in both the absence and presence of inducer. In the absence of inducer, (Table 4, “- MitoC”) the concentration of attP copies in culture supernatants of C. rodentium ΔrecA (Φstx2dact), predicted to be defective for SOS induction, did not increase between t=0h and t=4h, with an average relative attP production of 0.5. Lysogens deficient in the antiterminator Q, required for late gene transcription, or deficient in the S and R endolysins, which promote the efficient release of phage from infected bacteria, were also deficient in relative attP production in the absence of inducer (Table 4). Finally, C. rodentium (Φstx2dacΔB), predicted to replicate but not package phage genomes, showed no defect in the production of attP copies in the culture supernatant in the absence of inducer, with relative phage production ratio of 4.1. However, as described below, DNAse sensitivity assays suggested that these attP sequences are likely not packaged into phage particles.
The mutants defective in baseline phage production were similarly defective in the titer of attP copies after induction with mitomycin C (Table 4, “+ Mito C”). Induction of C. rodentium ΔrecA (Φstx2dact) resulted in a relative attP production value of only 29, i.e. eight-fold lower than wild type. C. rodentium (Φstx2dactΔQ), and C. rodentium (Φstx2dactΔSR), each also demonstrated dramatically diminished attP copies in mitomycin C-induced culture supernatants, with relative attP production of approximately 6. Finally, C. rodentium (Φstx2dactΔB), generated wild type levels of phage genome copies, with a 209-fold increase in relative attP production. However, DNAse treatment of supernatants diminished this value more than 23-fold, whereas parallel treatment diminished the relative attP production by wild type C. rodentium (Φstx2dact) less than 1.5-fold (Table 4, “+Mito C + DNAse”), consistent with a defect in packaging of Φstx2dact genomes in the absence of the B portal protein.
Proteins required for the SOS response and/or late gene transcription are essential for Stx2dact production
To determine which host or phage functions are required for production of Stx2dact in vitro, we measured Stx2dact in culture supernatants by ELISA [53]. To quantitate non-induced levels of Stx, and to provide ample time for toxin to accumulate, we grew triplicate cultures of the C. rodentium (Φstx2dact) or the mutant derivatives described above for four hours (t=4h) beyond mid-log phase (defined as t=0h). Stx2dact was present in the culture supernatants of wild type C. rodentium (Φstx2dact) at approximately 50 ng/ml/OD600 unit, consistent with previous measurements [53]. Prophage excision and phage production were not required for this basal level of Stx2dact: culture supernatants of C. rodentium (Φstx2dactΔint), which did not harbor detectable phage (Table 4), contained equivalent amounts of toxin (Fig. 2A, “Δint”). Uninduced culture supertants of C. rodentium (Φstx2dactΔSR) contained levels of Stx2dact two-fold lower than (and statistically indistinguishable from) wild type, consistent with the moderately (5-fold) lower levels of phage found in cultures of wild type C. rodentium (Φstx2dact) (Table 4, “- MitoC”). Supernatants of C. rodentium (Φstx2dactΔB), which contained attP DNA but relatively few packaged phage (Table 4), also produced levels of Stx2dact statistically indistinguishable from wild type. Finally, in contrast, C. rodentium ΔrecA (Φstx2dact), which is unable to mount an SOS response, and C. rodentium (Φstx2dactΔQ), which cannot transcribe phage late genes, including stx2dactA and stx2dactB, were defective for basal levels of Stx2dact production (Fig. 2A, “ΔrecA”, “ΔQ”).
We also assessed Stx production by wild type C. rodentium (Φstx2dact) and mutant derivatives after 4h of mitomycin C induction. Given that mitomycin C-induced Φstx2dact functions may be involved in the release of toxin from the bacterial host [27], we assessed toxin in cell pellets and in culture supernatants separately. As previously observed [52], mitomycin C induction resulted in a more than 100-fold increase of Stx2dact in culture supernatants (Fig. 2B, “WT”). A nearly equivalent amount of toxin remained associated with the bacterial cell pellet, suggesting that under these conditions, a significant fraction of bacteria remained unlysed. Culture supernatants or cell pellets of the C. rodentium ΔrpoS (Φstx2dact) mutant predicted to be defective in the bacterial stress response, or the C. rodentium ΔqseC (Φstx2dact) mutant defective for quorum sensing, showed wild type levels of Stx2dact (Fig. S3), as did C. rodentium ΔqseF (Φstx2dact) (data not shown), indicating that neither the bacterial stress response nor the QseC- or QseF-mediated quorum responses were required for toxin production. Culture supernatants of C. rodentium (Φstx2dactΔint) and C. rodentium (Φstx2dacΔB), which showed no defect in basal levels of toxin production (Fig. 2A), also contained amounts of cell-associated toxin and supernatant-associated Stx2dact indistinguishable from wild type (Fig. 2B, “Δint” and “ ΔB”), despite the lack of prophage excision and/or phage production in these mutant strains. The ΔSR lysogen, defective for phage endolytic functions, produced wild type levels of cell-associated Stx2dact at 4 h postinduction, but supernatant-associated toxin was approximately ten-fold lower than wild type levels (Fig. 2B, “Δ SR”). This difference is consistent with a defect in bacterial lysis and Stx2dact release, but did not reach statistical significance. In addition, by 16 h post-induction of the ΔSR lysogen, Stx2dact was detected in supernatants at levels similar to that of the WT strain (Fig. 2C), suggesting that any defect in R and S proteins results in a delay rather than an absolute block in toxin release. Finally, however, deficiency in the RecA or Q proteins was associated with a near-complete absence of Stx2dact in cell pellets or supernatants (Fig. 2B, “ΔrecA” and “ ΔQ”), reinforcing the notion that these proteins, which are required for the SOS response and/or transcription of the stx2dact ([74] [66]) are essential for Stx2dact production.
C. rodentium (Φstx2dact) undergoes lytic induction during murine infection
Stx-encoding prophages undergo lytic induction during EHEC infection of germ-free or antibiotic-treated mice [40, 41, 70], and our comprehensive survey of prophage integration sites in fecal microbiota (Table 3) indicated that C. rodentium (Φstx2dact) undergoes some degree of lytic induction during infection of conventional mice. To assess this induction further, we infected conventionally raised C57BL/6 mice with C. rodentium (Φstx2dact) by oral gavage and measured fecal shedding of both the infecting strain, by plating for CFU, and Φstx2dact, by quantitating attP (non-integrated phage) copies by qPCR. As previously observed, by day 3 post-infection, C. rodentium (Φstx2dact) was detected in the stool at 8 ⨯ 107 per gram, and reached 9 ⨯ 1010 per gram by day 6 post-infection ([64]; Fig. 3, “CFU of WT”). Further, murine infection by this strain was indeed associated with lytic induction, as excised phage genomes were detected in stool at all time points (Fig. 3, “Phage from WT”).
Interestingly, given the relatively high phage production by C. rodentium (Φstx2dact) in vitro, the amount of phage detected in stool was quite low. At day 3 post-infection, 5 ⨯ 106 attP copies were detected per gram of stool, a value 16-fold lower than the concentration of viable C. rodentium (Φstx2dact) in stool at that time point. By day 6 post-infection, attP copies had increased to 5 ⨯ 107 per gram of feces, but were approximately 600-fold lower than the fecal bacterial counts. C. rodentium Φ(stx2dact) thus undergoes lytic induction and growth in this murine model, although not to the degree seen in vitro.
Lethal disease in mice correlates with the ability to produce Stx2dact but not with the ability to produce phage
To test the importance of SOS induction and phage functions on disease in our microbiota-replete model of infection, we infected C57BL/6 mice with C. rodentium (Φstx2dact) and mutant derivatives by oral gavage. The wild type and all mutant lysogens colonized mice similarly (Fig. S4). C. rodentium ΔrecA (Φstx2dact) and C. rodentium (Φstx2dactΔQ), the two mutant lysogens that displayed dramatic defects in basal and mitomycin C-induced levels of Stx2dact in vitro, were the only ones incapable of causing sickness or death (Fig. 3, “ΔrecA” and “ ΔQ”), supporting the hypothesis that induction of an SOS response and the subsequent expression of phage late genes, including stx genes, are required for Shiga toxin production during infection of a microbiota-replete host.
The RpoS-deficient and QseC-deficient C. rodentium (Φstx2dact) mutants that are compromised in bacterial stress and quorum-sensing responses, respectively, retained the ability to cause weight loss and lethality with kinetics that were indistinguishable from that of WT C. rodentium(Φstx2dact) (Fig. S5). Thus, although previous results indicated that some quorum sensing mutants display diminished virulence during infection by non-Stx-producing C. rodentium [75], our results are consistent with the the ability of these strains to produce wild type levels of Stx2 after SOS induction (Fig. S3). In addition, the lack of endolysins that appeared to somewhat delay release of Stx2dact into supernatants by C. rodentium (Φstx2dactΔSR) was not reflected by any delay in the kinetics of weight loss or lethality in infected mice (Fig. 3, “ΔSR”), consistent with the ability of this strain to produce wild type levels of Stx2dact upon extended culture in vitro.
Finally, the production of intact phage is not essential to disease in this model. C. rodentium (Φstx2dactΔB), which is unable to generate intact phage in vitro, and C. rodentium (Φstx2dactΔint), which can neither generate excised phage genomes in vitro or in vivo, both retained full virulence in this model. We conclude that in this microbiota-replete model of EHEC infection, disease progression correlates exclusively with the ability to produce Stx2, regardless of the lysogen’s ability to amplify the stx2 gene by phage excision and genome amplification, or by the production of phage that are capable of secondary infection of commensal bacteria.
Discussion
Commensal organisms have the potential to suppress or enhance phage induction and Stx production. Although a role for induction of stx-encoding prophages in the production of Stx and serious disease during animal infection has been well documented in antibiotic-treated and germ-free mice [40, 41, 70], we used a murine model of EHEC infection that features an intact microbiome.
To investigate phage functions required for C. rodentium (Φstx2dact) to produce Stx and cause disease in conventional mice, we first characterized prophage genetic structure. Φstx2dact prophage was integrated into the C. rodentium dusA gene, an integration site utilized by prophages in over 200 bacterial species [67]. Although the orientation of the regulatory and late genes within the Φstx2dact prophage is noncanonical with respect to attL and attR (with int adjacent to attL; Fig. 1), this orientation has been previously observed in at least one other lambdoid phage. In addition, Φstx2dact genes encoding several key phage proteins were identified by homology, and their inactivation had the predicted effects on phage development and production (Table 4; [73]). For example, antiterminator Q and integrase were required for phage production, as measured by detection of attP, and portal protein B appeared to be required for packaging of phage DNA into DNAse-resistant virions.
Stx production in vitro by the prophage mutants, as well as by a host recA mutant, confirmed that prophage induction, i.e., the SOS-dependent process required to initiate a temporal program of phage gene expression that normally leads to phage lytic growth, is essential for high-level Stx2 production in vitro. Mitomycin C treatment of C. rodentium (Φstx2dact) resulted in a greater than 100-fold increase in Stx2dact in culture supernatants, similar to the mitomycin C-mediated increase in Shiga toxin production by EHEC ([41]; Fig. 2). Two signaling pathways, mediated by RpoS and QseC, previously demonstrated to influence SOS induction of EHEC in vitro, had no effect on Stx2dact production by C. rodentium (Φstx2dact). In contrast, and as expected, RecA, required for mounting an SOS response, was necessary for this enhanced production of Stx2dact (Fig. 2). It was previously shown that inactivation of the EHEC prophage repressor CI, a key step in the SOS response, is required for the increase in EHEC Stx production upon mitomycin C induction in vitro [41].
Despite the previous observation that the increase in phage genome copy number plays the most quantitatively important role in mitomycin C-enhanced Stx1 production by Stx phage H-19B [27], we found that integrase-deficient C. rodentium (Φstx2dact), which is deficient in phage excision and replication (Table 4; [72]), produced levels of Stx2dact indistinguishable from wild type (Fig. 2). Thus, the ability to cause lethal Stx2-mediated disease was not correlated with the ability to generate infectious phage. Apparently, enhanced expression of late genes stx2dactA and stx2dacB still occurs in the absence of integrase and is sufficient for wild type levels of Stx2dact production. As expected, antiterminator protein Q, required for the transcription of late genes including stx, was essential for Stx2dact production by C. rodentium (Φstx2dact), consistent with previous findings for the Stx2 phage Φ361 [26]. Finally, the S endolysin of Stx phage H-19B was previously shown to promote the timely release of toxin after mitomycin C induction [27]; we found that deficiency of the RS endolysins encoded by Φstx2dact appeared to diminish the release of Stx2dact into culture supernatants at 4 hours post-induction (Fig. 2B). However, the decrease was not statistically significant, and RS-deficiency had no discernible effect on toxin release by 16 hours (Fig. 2C).
Whereas previous work in streptomycin-treated or gnotobiotic murine models has demonstrated that induction of the lytic developmental program of Stx phage occurs during infection and is required for disease [40, 41, 70], we document here that prophage induction occurs and is critical for productive infection of mice with intact microbiota. Our evidence includes first that attP sequences (indicative of excised, uningegrated phage genomes) were detected in the feces of infected mice, as revealed by deep sequencing (Table 3), or by qPCR (Fig. 3). Second, the ability of C. rodentium (Φstx2dact) mutant derivatives to trigger the lytic cycle in vitro upon induction, with concomitant expression of the late stx2dact genes, correlated perfectly with the ability to cause lethal disease in mice. C. rodentium (Φstx2dact) derivatives deficient in phage integrase, portal protein B, or host regulators RpoS or QseC were entirely competent for production of high levels of Stx2dact (Figs. 2 and S3), and upon infection of mice, each of these mutants retained the ability to cause weight loss and death, with kinetics indistinguishable from the wild type lysogen (Figs. 4 and S5). In contrast, prophage gene Q, critical for late gene transcription, was required for both in vitro Stx2 production and lethal infection. RecA, essential for the initiation of the SOS response that leads to prophage induction, was required for lethality after oral inoculation of C. rodentium (Φstx2dact), consistent with the previous finding that RecA was required for lethality following intravenous EHEC infection of conventional mice [76]. Interestingly, human neutrophils and hydrogen peroxide are capable of increasing Stx production by EHEC in vitro, perhaps by triggering oxidative damage [77]. Our use of conventional C57BL/6 mice, for which many mutants are available, may facilitate studies to determine whether a specific inducing stimulus is responsible for lytic induction of C. rodentium (Φstx2dact).
We detected more than 1 ⨯ 109 phage/ml in uninduced mid-log cultures, suggesting that there is a high level of spontaneous induction under in vitro culture conditions. In contrast, despite severe Stx2dact-mediated disease manifestations during productive infection by C. rodentium (Φstx2dact), the number of attP sequences detected in feces was extremely low, suggesting that the level of prophage induction during infection may also be low. On day 6 post-infection, only 0.44% of all phage genomes detected were excised, compared to 99.66% that were integrated, reflecting intact prophage (Table 3). Depending on the day post-infection, excised phage detected by qPCR numbered 20- to 1000-fold fewer than viable C. rodentium (Φstx2dact) cells (Fig. 3). Notably, previous work using a genetic reporter to indicate activation of lytic promoters of EHEC Stx phage 933W showed that the intestinal environment of a gnotobiotic mouse was strongly inducing [41]. While we cannot rule out the possibility that the low number of Φstx2dact attP sequences detected in feces reflects an instability of phage particles or some other factor in the intestinal milieu, our findings are consistent with the possibility that a low rate of Φstx2dact induction may be sufficient to promote disease in this model, and that the presence or absence of the gut microbiota may have a consequential effect on disease outcome.
Consistent with the low level of phage detected in productively-infected mice, we found no evidence of Φstx2dact lysogeny of commensal bacteria during C. rodentium (Φstxdact) murine infection, suggesting that secondary infection of commensals by this phage is rare and that successive rounds of lytic infection are not an essential element of Stx production and disease in this microbiota-replete infection model. Given that the methods to measure phage particles utilized in this study can be applied to patient samples, future studies will focus on the extent of lytic induction of Stx phage during human infection, and how it may correlate with disease outcome.
Materials and Methods
Bacterial strains and plasmids
Strains and plasmids used in this study are listed in Table 1.
Phage Φstx2dact whole genome sequencing, assembly, and integration site determination
Genomic DNA was isolated from 5 ml of strain C. rodentium (Φstx2dact::kanR) (Table 1) growr overnight at 37°C in LB broth containing chloramphenicol (12.5 μg/ml) and kanamycin (25 μg/ml). DNA was extracted using a DNeasy kit (Qiagen), according to the manufacturer’s protocol for Gram negative bacteria. A library of this DNA was then constructed for Illumina sequencing using Illumina TruSeq DNA Sample Preparation Kit per the manufacturer’s instructions. Following sequencing, the bacterial genome was assembled de novo into 1500 contigs using assemblers ABySS [57], and Edena [58]. The Bowtie2 program [59] was then used to map the stx2 gene against this assembled genome and the contig containing this gene was identified. When aligned to the C. rodentium genome, a 69594-bp contig revealed a 47,343 bp prophage containing the stx2 gene and other phage lambda-like gene sequences inserted into the host dusA gene. (Although the C. rodentium dusA gene is interrupted by the prophage genome, a potentially functional dusA gene is reconstituted at the attL bacterial/phage DNA junction by fusion with a prophage-derived open reading frame that we term “ΦdusA’” in Fig. 1.) The prophage sequence was deposited in GenBank as Φ1720a-02, accession number KF030445.1.
Integration of the prophage in both C. rodentium (Φstx2dact) and C. rodentium (ΦΔstx2dact::kanR) into the host dusA gene was verified by PCR amplification of the attL and attR phage-host junctions using primers DusF/PhageR and DusR/PhageF, respectively (Table 2), then DNA sequencing of the amplified junctions. Subsequent whole genome sequencing of C. rodentium (Φstx2dact) and C. rodentium (ΦAstx2dact::kanR) showed that, with the exception of the presence of the Φstx2dact sequences, they are identical to C. rodentium ICC 168, also known as strain DBS100 (GenBank accession number NC_013716.1), and to each other (data not shown).
Phage Φstx2dact genome annotation
The Φstx2dact genome sequence was first annotated using the program RAST (http://rast.nmpdr.org/ [60]. The annotation was further refined by analyzing each open reading frame using the NCBI program MEGABLAST against the GenBank nucleotide database.
Characterization of phage and prophage sequences in murine stool by massively parallel sequencing and analysis
DNA was extracted from fecal samples of 5 infected sick mice at 6 days post-infection, according to the method of Yang et al. [61]. Twenty mg stool samples were suspended in 5 ml PBS, pH7.2, and centrifuged at 100 ⨯ g for 15 min at 4°C. The supernatant was centrifuged at 13,000 ⨯ g for 10 min at 4°C, and the resulting pellet was washed 3 times in 1.5 ml acetone, centrifuging at 13,000 ⨯ g for 10 min at 4°C after each wash step. Two hundred μl of 5% Chelex-100 (Bio-Rad) and 0.2 mg proteinase K were added to the pellet and the sample was incubated for 30 min at 56°C. After vortexing briefly, the sample was centrifuged at 10,000 ⨯ g for 5 min and the supernatant containing the DNA was harvested and stored.
To characterize bacteria that harbor the Φstx2dact prophage, we sequenced the bacterial bacterial-host attL prophage junction and adjacent bacterial DNA by following, with slight modifications (Suppl. Fig. 1), the methodology of Klein et al. [62] for constructing high-throughput sequencing libraries that contain a repetitive element (in this case, the phage int (integrase) gene). Briefly, genomic DNA was sheared by sonication to a size of 100-600 bp, followed by addition of ~20 deoxycytidine nucleotides to the 3’ ends of all molecules using Terminal deoxynucleotidyl Transferase (TdT, Fig. S1, step 2). Two rounds of PCR using a poly-C-specific and phage int gene-specific primer pair (PCR primers 1 and 2, Table 2) were used to amplify attL (Fig. S1, step 3) and to add on sequences necessary for high-throughput sequencing (PCR primers 3 and 4, Table 2, and Fig.S 1, step 4)).
Amplicons were sequenced using the MiSeq desktop sequencer (Ilumina) and primer Seq-P (Table 2), providing reads of up to 300 bp. As amplicons spanned the region from the phage int gene, through attL, and into the adjacent host genome (see Fig. 1B), reads of this length were required. 17,868,095 sequences encompassing 5 Gb were downloaded to the Galaxy server (https://usegalaxy.org/) and analyzed (Table 3). We first excluded sequences that clearly reflected attL (i.e., contained the 184 bp of ΦdusA’followed by C. rodentium dusA), indicating the prophage inserted into the C. rodentium genome. Of the remaining 801,959 sequences, 75,962 (0.44% of the total) encoded the intact attP site, implying that they were circular. These latter sequences presumably reflected excised circular phage genomes, possibly undergoing early theta DNA replication, ultimately leading to phage production. The remaining 725,997 sequences encoded only strings of A’s and/or C’s, and were eliminated from consideration.
Generation of C. rodentium (Φstx2dact) deletion constructs
Deletion mutants of C. rodentium (Φstx2dact) in the prophage or the host genome were generated using a modified version of a one-step PCR-based gene inactivation protocol [63, 64]. Briefly, a PCR product of the zeocin-resistance gene and its promoter region flanked by 70-500 bp homology of the region upstream and downstream of the targeted gene was generated using the primers listed in Table 2. The chromosomal DNA served as template when the flanking regions were 500 bp in length on either side of the Zeocin cassette. The PCR product was electroporated into competent C. rodentium (Φstx2dact) cells containing the lambda red plasmid pKD46 and recombinants were selected on plates containing chloramphenicol and zeocin (75 μg/ml). Replacement of the gene of interest with the zeocin resistance cassette was confirmed using specific primers (Table 2). At least two independent clones, validated using PCR, were obtained and subsequently analyzed.
Quantification of Stx2 produced in vitro
Overnight 37°C cultures of C. rodentium (Φstx2dact) or deletion derivatives were diluted 1:25 into 10 ml of fresh medium with appropriate antibiotics. Two independently derived clones for each mutant were tested, with indistinguishable results. The cultures were grown at 37°C with aeration to an OD600 of 0.4, and one ml of each culture was set aside (Table 4, “t=0h”) The remaining culture was split into 2 cultures. These cultures were grown for a further 4 hours (Table 4, “t=4h”) in the absence or presence of 0.25 μg/ml mitomycin C. (We first measured phage and Stx2 production at various times post-induction and found the 4-hour time point to be optimal for obtaining maximal phage and Stx2 following mitomycin C induction). Culture pellets and supernatants were then harvested by centrifugation at 17,800 ⨯ g for 5 minutes at room temperature. For C. rodentium (Φstx2dact) and C. rodentium (Φstx2dactΔSR), a portion of each culture was also collected after ~16 h of incubation (“t=16h”). Supernatants and pellets were quantitated for Stx2 by ELISA, as described previously [52].
Mouse infection studies
Mice were purchased from Jackson Laboratories and maintained in the Tufts University animal facility. All procedures were performed in compliance with Tufts University IACUC protocols. Seven to eight-week-old female C57BL/6J mice were gavaged with PBS or ~5⨯108 CFU of overnight culture of C. rodentium (Φstx2dact) or deletion derivatives in 100 μl PBS. Inoculum concentrations were confirmed by serial dilution plating. Fecal shedding was determined by plating dilutions of fecal slurry on either chloramphenicol, to detect wild type C. rodentium (Φstx2dact), or chloramphenicol-zeocin plates, to detect deletion derivatives marked with a zeomycin resistance gene [52]. Body weights were monitored daily, and mice were euthanized upon losing >15% of their body weight.
DNA from infected mice fecal pellets was isolated using the QIAGEN DNeasy Blood and Tissue kit with modifications. Fecal pellets were incubated with buffer ATL and proteinase K overnight at 55°C. Buffer AL was added, and after mixing, pellets were further incubated at 56°C for 1 h. Pellet mixtures were then centrifuged at 8000 rpm for 1 min and the pellets were discarded. Ethanol was added to the supernatants, which were processed according to the manufacturer’s protocol. DNA concentrations were determined using a NanoDrop™ spectrophotometer. qPCR was performed as described below.
Quantification of phage genomes by qPCR
Excised phage genomes in cell supernatants were quantitated by qPCR. Supernatants were serially diluted 1:10, 1:100 and 1:1000 in distilled water. Separate reactions using two μl of the various dilutions as a template were carried out in duplicate. qPCR master-mix (Bio-Rad) was prepared according to the manufacturer’s instructions, using the attP primer set (Table 2) to detect copies of excised phage DNA. Results were compared to a standard curve, derived from a known concentration of a template fragments generated from amplifying C. rodentium (Φstx2dac) DNA using attP primers. The template was serially diluted, in duplicate, to detect copy numbers ranging from 1010 to 102. qPCR reactions were carried out as follows: 95°C for 3 min, followed by 35 cycles of 95°C for 1 min, 58°C for 30 sec, and 72°C for 1 min.
Statistical tests
Data were analyzed using GraphPad Prism software. Comparison of multiple groups were performed using the Kruskal-Wallis test with Dunn’s multiple comparison post-test, or 2-way ANOVA with Bonferroni’s post-tests. In all tests, P values below 0.05 were considered statistically significant. Data represent the mean ± SEM in all graphs.
Acknowledgements
We are grateful to Sara Roggensack for suggesting the PhageSeq methodology, David Lazinski for help with preparation of PhageSeq libraries, Martin Marinus for critical review of the manuscript, and Michael Pereira, Martin Marinus, and members of the Leong Lab for many useful suggestions.
Footnotes
1 Grant support: This work was supported by NIH grants R21AI107587 and 2R01AI046454 to JML.