Shigella clinical isolates used in this study are listed in Table 1. S. flexneri strains were grown in Tryptic Soy Broth (BD Difco, Franklin Lakes, NJ, USA) at 37°C with aeration or on Tryptic Soy Broth plates containing 1.5% agar and 0.025% Congo red. E. coli K-12 strain MG1655 was grown in Luria-Bertani broth and on Luria-Bertani agar plates. Kanamycin and ampicillin were used at 50 μg/mL and 100 μg/mL, respectively.

DNA lysates were used for PCR for stx_1a with primers stx1-det-F1 and stx1-seq-R1, and for stx₂ with primers stx_F/stxR1. All strains were verified to contain the virulence plasmid of S. flexneri by PCR amplification of virF with primers VirF1 and VirF2.

To show that stx_1a was phage encoded, we analyzed lysates by PCR with primer pairs Stx1R2/Phage_stxR2 and Phage_stx1F2/Stx1F2. The insertion site of the phage into S. flexneri locus S1742 was determined by amplifying the upstream region of S1742 and an early phage gene with primers S1742_up/Stx_phage_up and by amplifying a late phage gene and the downstream region of S1742 with primers Stx_phage_dn/S1742_dn. All PCRs were conducted by using PCR Master Mix (2X) (Fermentas, Pittsburgh, PA, USA) according to the manufacturer’s specifications. Primer sequences and expected amplicons are listed in Table 2.

recA was replaced with a kanamycin-resistance cassette by using λ red recombination (15). Primers RecAko-site1 and RecAko-site2 were used to amplify kan from pKD4 with 5′ and 3′ overhangs homologous to internal regions of recA. Kanamycin-resistant colonies of BS766 (15) were double purified and screened by PCR with primers RecA-1/RecA-3 for detection of the size difference between chromosomal recA and the kanamycin-resistance cassette. This mutant was used as the donor for growing a P1L4 lysate, which was used to transduce the recA::kan mutation into BS937, BS938, and BS974. Kanamycin-resistant transductants were purified and confirmed by use of PCR, as described above.

The cytotoxicity of bacterial samples for Vero cells was determined as previously described (16,17). In brief, 100 μL of diluted samples was overlain in 96-well plates containing confluent monolayers of Vero cells and incubated for 48 hours at 37°C in 5% CO₂. Viable cells were fixed with 10% formalin and stained with 0.13% crystal violet. The optical density (OD) of the stained wells was measured at 630 nm by using a BioTek (Winooski, VT, USA) EL800 spectrophotometric plate reader. The CD₅₀ (cytotoxic dose that kills 50% of the cells) was calculated by determining the inverse dilution of the bacterial sample that was required to kill 50% of the Vero cells.

Overnight supernatants were serially diluted 10-fold in medium. We mixed 100 μL of diluted samples with 100 μL medium, a 1:25 dilution of F45 polyclonal anti-Stx1/Stx1a antiserum (17,18), or a 1:25 dilution of rabbit polyclonal antiserum against S. flexneri whole cell lysate. Toxin samples were incubated with antibody for 2 hours at 37°C in 5% CO₂. We then applied 100 μL of the toxin–antibody mixture to Vero cells and incubated as above.

Overnight cultures of bacteria were inoculated 1:100 into Tryptic Soy Broth, and 2 hours after inoculation a final volume of 0.5 μg/mL mitomycin C (Sigma, St. Louis, MO, USA) was added to the cultures. To monitor the induction of bacterial lysis, we read the OD₆₀₀ hourly over a period of 8 hours. Induction of bacterial lysis was noted as a 3-4–fold decrease in OD₆₀₀ compared with the stx_1a-negative control strains.

To determine the effect of mitomycin C on production of Stx1a and prophage induction, we collected whole cell lysates and supernatants 3 hours after addition of mitomycin C. Samples were then analyzed for cytotoxicity on Vero cells. For isolation of phage particles, supernatants were prepared similarly, except that a final concentration of 10 mmol/L MgSO₄ and a drop of chloroform were added after centrifugation. As described previously, 100 μL of phage lysate was absorbed onto 200 μL of E. coli MG1655 for 20 minutes at 37°C (19). Molten L-agar top agar (19) containing 10 mmol/L MgSO₄ was added to the phage/bacteria mixture and poured onto L-agar plates. Plates were incubated overnight at 37°C, and plaque-forming units (PFUs) were counted.

Supernatants containing phage were prepared from mitomycin C–induced culture of BS937. Phage lysate was spotted onto an L-soft agar overlay containing either E. coli MG1655 or S. flexneri 2457T and incubated overnight at 37°C. A loop from the zone of clearing was streaked for isolation of single colonies, which were subsequently screened for stx_1a genes by PCR. Positive colonies from the initial screening were double colony purified, and PCR was repeated to ensure that the colonies were positive for stx_1a. MG1655 lysogens were confirmed to not be contaminates of the donor strain, BS937, by testing for S. flexneri chromosomal and virulence plasmid genes by use of PCR primer pairs IpaHF/IpaHR and VirB01/IpaFWD (Table 2), respectively. Similarly, 2457T lysogens were analyzed by PCR as described above. Two independently isolated lysogens of MG1655 and 2457T were used for further analysis.

Virulence-associated phenotypes were determined by a gentamicin protection invasion assay in HeLa cells and by plaque formation in L2 monolayers, as previously described (20,21). Both assays were conducted 3 independent times and included technical duplicates or triplicates in each individual experiment.

DNA was isolated from overnight cultures by using a QIAGEN DNEasy Kit (Valencia, CA, USA). Samples were prepared for sequencing by using a Nextera XT DNA Sample Preparation Kit (Illumina, San Diego, CA, USA) and sequenced on an Illumina MiSeq sequencing system. The phage sequence was assembled by mapping the reads to the reference phage NC_004913.2 by using Bowtie 2 version 2.1.0 (http://sourceforge.net/projects/bowtie-bio/files/bowtie2/2.1.0/).

The complete phage sequence of ϕPOC-J13 from strain BS937 was submitted to GenBank. The sequence is available under accession no. KJ603229.

BS937 and BS938 (Table 1) were acquired from the Hawaii and Massachusetts state laboratories, which had determined the isolates to be positive for stx_1a by PCR. Both strains shared the same pulsed-field gel electrophoresis (PFGE) XbaI pattern, JZXX01.0357, as indicated in the Centers for Disease Control and Prevention PulseNet database (http://www.cdc.gov/pulsenet). To identify other clinical isolates of S. flexneri 2 that might be stx_1a-positive, we searched the PulseNet database for strains that matched this PFGE pattern. From state public health laboratories, we obtained 18 additional strains of S. flexneri that matched this PFGE pattern. We also received time-matched, but not PFGE-matched, strains of S. flexneri as negative controls. Six additional strains that had been confirmed to be stx_1a-positive S. flexneri were acquired from the Centers for Disease Control and Prevention. Clinical strains included in this study and their sources are listed in Table 1.

The stx_1a-positive S. flexneri strains had been isolated over 13 years (2001–2013). They were isolated during all months except August, indicating that seasonality is not involved in the emergence and/or spread of stx_1a-encoding S. flexneri. Among patients from whom stx_1a-positive S. flexneri strains were isolated, no incidences of HUS were reported, suggesting that the stx_1a-positive S. flexneri strains did not cause more severe disease than would typically be caused by such strains lacking stx_1a. Among 22 patients for whom travel information was available, 9 reported no foreign travel or knowledge of contact with persons who had traveled. The 13 patients who reported travel had all recently visited Hispañiola or interacted with a traveler who had returned from this region (Table 1).

The presence of stx_1a in the S. flexneri strains was confirmed by PCR (data not shown). All isolates that matched PFGE pattern JZXX01.0357 or had previously been shown to encode stx_1a yielded a PCR product of the correct size for the toxin gene. PCR analysis for stx₂ did not produce a product. The 3 negative controls did not generate a PCR product for either stx_1a or stx₂.

To determine if Stx1a was released from the bacteria, we tested supernatants from overnight cultures in a Vero cell cytotoxicity assay. All stx_1a-positive isolates released a toxin into the supernatant, which killed Vero cells. To confirm that the toxin responsible for killing the cells was Stx1a, we tested overnight supernatants from 3 representative isolates for Vero cell cytotoxicity after neutralization by anti-Stx/Stx1a antiserum. After neutralization, supernatants were no longer cytotoxic to Vero cells (Figure 1). These findings demonstrate that the extracellular product responsible for cytotoxicity to Vero cells is indeed Stx1a and not a different protein being released by the stx_1a-expressing S. flexneri.

stx is generally found encoded by functional prophages (22). The prophage lytic cycle can be induced with DNA damaging agents (23); therefore, to address whether the toxin carried in stx_1a-encoding S. flexneri was associated with a prophage, we tested sensitivity to lysis when grown in the presence of mitomycin C. All Stx1a-producing S. flexneri isolates showed a sharp decrease in OD₆₀₀ within 3–4 hours after addition of mitomycin C, whereas the stx_1a-negative strains showed no decrease in OD₆₀₀ (data not shown).

Because all stx_1a-encoding S. flexneri isolates behaved similarly in the assays, we selected 3 isolates (BS937, BS938, and BS974) to characterize more thoroughly. To further investigate the response to mitomycin C, we grew log-phase cultures and collected samples to measure cell-associated and released toxin. Supernatants and whole cell lysates from bacteria treated with mitomycin C exhibited elevated cytotoxicity to Vero cells (Figure 2). To determine if Stx1a–producing S. flexneri generated infectious phage, we analyzed supernatants from untreated and mitomycin C–treated cultures in a PFU assay. After treatment with mitomycin C, PFUs in supernatants of BS937, BS938, and BS974 increased ≈1,000-fold (Figure 3).

Induction of the prophage lytic cycle by mitomycin C is caused by the SOS response (24). During the SOS response, the bacterial protease RecA becomes active and cleaves the phage repressor cI, which maintains the phage in a quiescent state under non-SOS conditions. Cleavage of cI enables transcription of the phage antiterminator Q, which activates the late phage genes, including stx (25). STEC recA mutants no longer release toxin or respond to agents that trigger the SOS response (24,26,27).

To investigate whether similar regulation occurs in Stx1a-producing S. flexneri, we constructed recA deletions in BS937, BS938, and BS974. recA mutants were cultured with mitomycin C as above, and samples were collected to measure the presence of toxin and phage. In the absence of mitomycin C, the recA mutants produced and released toxin in amounts comparable to those of the parental strains; however, when cultured with mitomycin C, the recA mutants did not exhibit increased cytotoxicity to Vero cells (Figure 2). Additionally, in the absence or presence of mitomycin C, no PFUs were enumerated from the recA mutants. Collectively, these data suggest that stx_1a in S. flexneri is carried by a lambdoid prophage.

To identify the location of stx_1a, we sequenced BS937, BS938, and BS974. Whole-genome sequencing confirmed that stx_1a was encoded within a 62-kb lambdoid prophage. To extend the analysis to all the clinical isolates in this study, we used the PCR strategy and primer design shown in Figure 4. To ensure that stx_1a in all isolates was phage encoded, we designed primers to amplify from stxA_1a and stxB_1a (encoding subunits A and B of Stx1a) and 1 kb either upstream or downstream of the stx_1a operon. All stx_1a-positive strains yielded a PCR product consistent with the toxin being phage encoded. No DNA was amplified from the stx_1a-negative isolates.

Whole-genome sequencing revealed that the phage was inserted into locus S1742 (which encodes a putative oxidoreductase) of the S. flexneri chromosome. Primers were designed (Figure 4) to determine if the stx_1a-encoding phage inserted into S1742 for all isolates. All stx_1a-positive strains generated the expected PCR product when amplified with primers specific for the early phage sequence and upstream of S1742 and with primers directed to the late phage sequence and downstream of S1742. None of the stx_1a-negative strains yielded an amplified product with the primer pairs. A representative gel of the 4 amplifications for 1 stx_1a-negative and 6 stx_1a-positive isolates is shown in Figure 5. These data suggest that a phage has integrated into all stx_1a-positive isolates. We named this stx_1a-encoding phage ϕPOC-J13.

E. coli MG1655 and S. flexneri 2457T were lysogenized as described earlier. To test for production and release of Stx1a, we examined whole cell lysates and supernatants from overnight cultures of MG1655 and 2457T lysogens in Vero cell cytotoxicity assays. The average CD₅₀/mL of supernatants from MG1655 and 2457T lysogens was 1 ×10⁵, similar to that of the stx_1a-positive S. flexneri clinical isolates. Lysogens were also tested for the presence of phage in supernatants of overnight cultures by determining PFUs. The 2457T lysogens released ≈10⁶ PFUs/mL; however, no viable phage could be recovered from MG1655 lysogens.

To confirm that stx_1a in the lysogens was phage associated, we conducted PCR amplification as described above (Figure 4). The MG1655 and 2457T lysogens yielded a PCR product that indicated that the phage regions upstream and downstream of stx_1a were present. Lysogens were also tested for integration of the phage into locus S1742 by use of the primers illustrated in Figure 4. In 2457T lysogens, ϕPOC-J13 inserted into S1742 and produced a PCR product of the expected size. Similarly, in lysogens of MG1655, ϕPOC-J13 inserted into ynfG, the E. coli S1742 homologue. These findings demonstrate that the integration of ϕPOC-J13 is site specific, as has been shown for other lambdoid prophages (9).

We wanted to determine if the presence of ϕPOC-J13 altered the virulence phenotypes associated with S. flexneri. Invasion and plaquing efficiencies of BS937, BS938, and BS974 were compared with those of laboratory strain 2457T (Table 3). BS938 and BS974 exhibited similar invasion efficiency as 2457T; however, invasion with BS937 was significantly higher. All 3 Stx1a–producing S. flexneri isolates showed plaquing efficiency comparable to that of 2457T, and the plaque diameters were consistent among all strains. Although it is unclear why BS937 was more invasive, the comparable level of cell-to-cell spread suggests that ϕPOC-J13 does not appreciably alter the virulence properties of S. flexneri.