Microbial Hazards and Emerging Issues Associated with Produce † A Preliminary Report to the National Advisory Committee on Microbiologic Criteria for Foods.

In the past two decades, the consumption of fresh fruits and vegetables in the United States has increased, and the geographic sources and distribution of fresh produce have expanded greatly. Concomitantly, public health officials have documented an increase in the number of reported produce-associated foodborne disease outbreaks in the United States. The Centers for Disease Control and Prevention (CDC) reports that the number of these outbreaks doubled between 1973 and 1987, and 1988 and 1991, and that the number of cases of illness associated with these outbreaks more than doubled. A variety of produce items have been affected. During 1995 alone, major outbreak investigations linked infections with Salmonella serotype Stanley to alfalfa sprouts, Salmonella Hartford to unpasteurized orange juice, Shigella spp. to lettuce and green onions, Escherichia coli O157:H7 to lettuce, and hepatitis A virus to tomatoes. In response to this apparent increase, the U.S. Food and Drug Administration asked the National Advisory Committee on Microbiological Criteria for Foods to address and better define the association of foodborne disease and microbial pathogens with fresh produce. A subcommittee formed in June 1995 is documenting relevant epidemiologic data, current industry practices, and laboratory data to identify potential hazards and related control strategies. This report presents the preliminary findings of that subcommittee.

Molecular subtyping of toxigenic Vibrio cholerae O139 causing epidemic cholera in India and Bangladesh, 1992-1993.

Since October 1992, > 150,000 cases of cholera have been reported from India and Bangladesh; the great majority of Vibrio cholerae isolates belong to the newly established serogroup O139. To better understand the interaction of genetic and epidemiologic factors responsible for their sudden appearance and rapid spread, representative toxigenic V. cholerae O139 isolates were molecularly characterized and compared with a set of toxigenic V. cholerae O1 and non-O1/non-O139 strains. DNA sequences of the cholera toxin B subunit gene and multilocus enzyme electrophoresis markers of V. cholerae O139 strains were identical to those of V. cholerae O1 isolates of the seventh pandemic. Two distinct ribotypes and four pulsed-field gel electrophoretic patterns were observed for O139 strains. V. cholerae O139 strains were very similar to V. cholerae O1 strains of the seventh pandemic but clearly different from the toxigenic V. cholerae strains of serogroups other than O1 and O139.

Epidemiologic application of a standardized ribotype scheme for Vibrio cholerae O1.

A standardized scheme of 27 different BglI ribotypes and subtypes of Vibrio cholerae O1 strains is proposed on the basis of data from 214 human and environmental strains isolated in 35 countries and 14 U.S. states over the past 60 years. The ribotype patterns obtained are reproducible and stable over time. Seven different but very similar ribotypes (1a to 1g) were observed among 16 strains of the classical biotype. Twenty ribotypes and subtypes were identified among 198 V. cholerae O1 strains of the El Tor biotype. Six different patterns were found among the strains causing the current seventh pandemic. Strains of ribotype 8 originated only in central African countries, while those of ribotype 3 originated mainly in Asia and the Pacific Islands. The most widely distributed strains were those of ribotype 6, which was subdivided into three very similar but still distinguishable subtypes. The present Latin American epidemic is caused by strains of ribotype 5. Strains of this ribotype were isolated from several other geographic locations but can be differentiated from the Latin American strains by other molecular methods. Strains associated with two documented environmental reservoirs exhibited three distinct ribotype patterns; those isolated from patients who ate food from the U.S. Gulf waters were all of ribotype 2, while the strains related to the northeast Australian rivers were of ribotypes 9 and 10. Nontoxigenic V. cholerae O1 strains originating in Latin America and the U.S. Gulf Coast did not form a specific cluster of ribotypes. Ribotyping in combination with other well-defined methods can assist in epidemiologic investigations, helping to trace the movement of strains and to identify their geographic origins.

Use of polymerase chain reaction for detection of toxigenic Vibrio cholerae O1 strains from the Latin American cholera epidemic.

In January 1991, an outbreak of cholera started in Peru and spread throughout most of Latin America within 8 months. As of March 1992, over 450,000 cases and approximately 4,000 deaths have been reported to the Pan American Health Organization. The causative organism is toxigenic Vibrio cholerae O1 of the El Tor biotype and is distinct from the U.S. Gulf Coast strains. A polymerase chain reaction (PCR) that amplifies a 564-bp fragment of the cholera toxin A subunit gene (ctxA) was used to identify toxigenic V. cholerae O1 strains. A total of 150 V. cholerae O1 isolates were tested. They were of unknown toxin status, were associated with recent outbreaks, and were isolated from patients, food, and water. One hundred forty isolates were found to be toxigenic both by PCR and the routine diagnostic enzyme-linked immunosorbent assay. Thirty-eight known toxigenic strains isolated worldwide from 1921 to 1991 were also positive in the PCR. A collection of 18 nontoxigenic V. cholerae O1 strains, 35 Escherichia coli heat-labile-enterotoxin-I-producing strains, 26 Campylobacter strains, and 8 strains of Aeromonas hydrophila, previously reported to produce cholera toxin-like toxin, were all negative in the ctxA PCR. We conclude that this PCR is a diagnostic method that specifically detects toxin genes in V. cholerae O1 strains in a reference laboratory. It is more rapid and less cumbersome than other diagnostic methods for detection of toxicity in these strains.

Enteropathogenicity of non-toxigenic Vibrio cholerae O1 for adult mice.

The enteropathogenic potential of 32 Vibrio cholerae O1 isolates that do not produce cholera toxin was examined in the orally inoculated, sealed adult mouse model. Live cultures (2 x 10(10) cfu/ml) of 7/16 clinical and 6/16 environmental isolates produced a positive intestinal fluid accumulation (FA) ratio that reached near maximum at approximately 5 h post-inoculation. Colony hybridization did not detect genes for cholera toxin, Escherichia coli heat-labile and heat-stable toxins, or shiga-like toxins. FA activity did not correlate precisely with cytotoxic activities on Chinese hamster ovary (28/32 positive), Vero (29/32) or HeLa (25/32) cells. Certain clinical and environmental isolates of non-toxigenic V. cholerae O1 appear to be enteropathogenic for the mouse, providing evidence that they may have pathogenic potential for humans through an as yet undefined mechanism(s).

A nested PCR followed by magnetic separation of amplified fragments for detection of Escherichia coli Shiga-like toxin genes.

The Shiga-like toxin (SLT) I and II genes in cytotoxic Escherichia coli strains were detected using a polymerase chain reaction (PCR) procedure. Identification and differentiation of SLT I and II was carried out using primers giving PCR-generated DNA fragments of different size for the two cytotoxins. A two-step PCR procedure utilizing three primers in a nested configuration for both SLT I and II was combined with magnetic separation to identify the toxin genes in a rapid, specific and sensitive test system designated DIANA (Detection of Immobilized Amplified Nucleic Acid). The first PCR was carried out using standard methods, and the product generated was used as primer in the second PCR. In this procedure one of the primers from the first PCR was used with biotin label, and the second (inner) primer was 32P-labelled. The double-stranded DNA fragments generated containing the two primers, were biotinylated on one 5' end and 32P-labelled on the other 5' end. These fragments were separated from the solution using streptavidin-coated super-paramagnetic microscopic beads. The test could detect and differentiate between SLT I and II in a positive/negative ratio of more than 20. The assay could detect five SLT-positive E. coli organisms in the 5 microliters test sample. The presence of 100-fold more SLT-negative strains in a sample did not adversely affect the test signal.

Genetic diversity among toxigenic and nontoxigenic Vibrio cholerae O1 isolated from the Western Hemisphere.

Multilocus enzyme electrophoresis was used to examine genetic relationships among and between toxigenic and non-toxigenic isolates of Vibrio cholerae O1 obtained from patients and the environment in the US Gulf Coast and surrounding areas. A total of 23 toxigenic and 23 non-toxigenic strains were examined. All the toxigenic and 7 of the non-toxigenic strains had the same alleles at 16 enzyme loci, whereas the balance of the nontoxigenic strains had 9 distinct combinations of alleles. This study suggests that all of the toxigenic strains belong to a single clone, and that while some of the non-toxigenic isolates were related, most were of diverse origin.

Laboratory observations on Plesiomonas shigelloides strains isolated from children with diarrhea in Peru.

Eleven strains of Plesiomonas shigelloides isolated from 10 Peruvian children with diarrhea were examined. All the strains were resistant to two or more antibiotics, most commonly ampicillin, gentamicin, erythromycin, kanamycin, and streptomycin. The strains were all negative in the Sereny and cell culture assays used to test for enteroinvasiveness. One strain showed cytotoxic activity on Vero cells. The strains showed no antigenic relationship with Shigella organisms. Both bioassays and enzyme-linked immunosorbent assays used for detection of Escherichia coli enterotoxins were negative. Nucleic acid probes for such toxins likewise gave negative results. The strains all possessed a large (approximately 200-megadalton) plasmid in addition to one or more other plasmids. Several different plasmid profiles were observed among these 11 P. shigelloides strains, indicating that the isolates were not acquired from a common source or from a single bacterial clone.

Colonization of neonates in a nursery ward with enteropathogenic Escherichia coli and correlation to the clinical histories of the children.

Stool samples were examined from 30 preterm neonates admitted to a nursery ward; 16 neonates had diarrhea, 12 constituted an age-matched control group without diarrhea, and 2 had an unknown history regarding diarrhea. Variable numbers of enteropathogenic Escherichia coli serotype O111:HNT strains possessing the gene coding for the enteroadherence factor (EAF) were found in stool samples from 13 of the neonates. No other microbiological enteropathogen was found. A total of 294 strains (9 or 10 from each neonate, comprising 229 E. coli and 65 Klebsiella pneumoniae strains) were characterized with respect to plasmid content and grouped into 37 plasmid profile groups. Diarrhea was found not to be correlated with any specific plasmid profile or with the presence of the EAF-positive strains but rather with the number of strains with one specific plasmid profile or with the number of EAF-positive strains (of the 9 or 10 strains) isolated from each stool sample. All the neonates who died had diarrhea (5 died of 16 with diarrhea); all five of the neonates who died possessed strains with one specific plasmid profile group, and EAF-positive strains were isolated from four of them. Of the seven neonates from whom seven or more EAF-positive isolates were isolated, three died, compared with only one of five of those from whom only a few (1 to 3 of 10) EAF-positive strains were isolated. Both plasmid profiling and genetic probing with the EAF probe were found to be good alternatives when serotyping is not available for identification of O111:HNT enteropathogenic E. coli strains.

Enteropathogenic Escherichia coli serotype O111:HNT isolated from preterm neonates in Nairobi, Kenya.

This investigation was initiated as a consequence of several cases of diarrhea in a nursery ward for preterm babies in Nairobi, Kenya. Ten lactose-positive colonies were isolated from the stools of each of 30 neonates, regardless of whether they had diarrhea; 229 strains were identified as Escherichia coli and 65 strains were identified as Klebsiella pneumoniae. Six strains were lost during laboratory handling. No other bacterial, viral, or parasitic enteropathogens were identified. Using synthetic alkaline phosphatase-labeled probes, the bacterial isolates were found to be negative for the presence of genes coding for heat-stable and heat-labile enterotoxins. Seventy-eight E. coli strains isolated from a total of 13 neonates possessed the E. coli enteropathogenic adhesion factor (EAF) gene, as demonstrated by the use of a cloned radiolabeled DNA fragment probe. These strains possessed similar plasmid profiles constituting a core plasmid profile, and while all adhered to HeLa cells, none produced Vero cell cytotoxins. The EAF gene was located on a 65-megadalton plasmid. Serotyping showed the strains to be of serogroup O111 and serotype H nontypable, a well known enteropathogenic type. Five neonates died during the outbreak, and the fatality rate was 30.7% (4 of 13) for neonates infected with EAF-positive E. coli strains compared with 7.7% (1 of 13) for neonates from whom only EAF-negative E. coli strains were isolated. K. pneumoniae only was isolated from five neonates.

Studies in gnotobiotic piglets on non-O157:H7 Escherichia coli serotypes isolated from patients with hemorrhagic colitis.

A number of verotoxin-producing Escherichia coli strains isolated from sporadic cases of hemorrhagic colitis in the United States over the last 5 yr were shown to belong to serogroups other than O157:H7-the serotype originally implicated in this disease. Experimental infection of gnotobiotic piglets with five such strains (0111:NM, 0145:NM, 045:H2, 04:NM, and Ound:NM) caused diarrhea resulting from mucosal lesions in the cecum and colon that were indistinguishable from those previously described in piglets infected with E. coli O157:H7. This suggests that, as with other categories of pathogenic E. coli, several serotypes cause hemorrhagic colitis in humans. The five E. coli strains that were compared with one O157:H7 strain and with an enteropathogenic calf strain (serotype 05:NM) caused a spectrum of disease ranging from moderate diarrhea (O157:H7) to severe illness (including septicemia and death) (0111:NM). Characteristic lesions, which were identical for all seven pathogenic strains, included bacterial attachment, effacement of the microvillus border, and dissolution of the cell membranes of surface and glandular epithelium, resulting in complete cell destruction. Some piglets exhibited neurologic signs of convulsions and ataxia. It is concluded that a number of E. coli serotypes, in addition to O157:H7, fulfill the present limited criteria for enterohemorrhagic E. coli, which include association with hemorrhagic colitis, production of one or more verotoxins, possession of a large plasmid (50-70 megadaltons), and induction of distinct mucosal lesions in the large bowel of gnotobiotic piglets.

Role of a 60-megadalton plasmid and Shiga-like toxins in the pathogenesis of infection caused by enterohemorrhagic Escherichia coli O157:H7 in gnotobiotic piglets.

Enterohemorrhagic Escherichia coli (EHEC) of serotype O157:H7 has two putative virulence factors: (i) a fimbrial adhesin, specified by a 60-megadalton (MDa) plasmid, and (ii) bacteriophage-specified cytotoxin(s), known as Shiga-like toxin (SLT) or verotoxin. The contribution of these factors to the pathogenesis of EHEC-induced disease in gnotobiotic piglets was examined. The bacterial strains included the following: two EHEC strains and their corresponding plasmid-cured derivatives; another EHEC isolate and its derivative which had spontaneously lost the ability to produce SLT; one E. coli K-12 transconjugatant containing a 60-MDa plasmid from an EHEC strain; two K-12 strains into which an SLT-producing phage had been transduced (one of these strains also carried a 60-MDa EHEC-derived plasmid); and the parent K-12 strain. Each strain was fed to four piglets, which were observed for diarrhea and examined for development of characteristic mucosal lesions 3 or 5 days after inoculation. All 24 piglets inoculated with the three EHEC strains and their respective derivatives (two plasmid cured and one SLT negative) showed the typical mucosal lesions of bacterial attachment: effacement of microvillous border and cell membrane dissolution culminating in destruction of surface and glandular epithelium in the cecum and colon. No such lesions were observed in 12 piglets inoculated with three strains of E. coli K-12, including the strain which carried both the 60-MDa plasmid and a phage which specified production of SLT. Moderate to severe diarrhea was observed in 16 piglets inoculated with two EHEC strains and their derivatives (one plasmid cured and one SLT negative). The third EHEC strain and its plasmid-cured derivative produced fewer typical mucosal lesions and no diarrhea. The reason for the reduced virulence of this strain was not clear. These results demonstrate that neither the 60-MDa plasmid nor the capacity to produce SLT is essential for expression of virulence by E. coli O157:H7 in gnotobiotic piglets.

A DNA probe to identify enterohemorrhagic Escherichia coli of O157:H7 and other serotypes that cause hemorrhagic colitis and hemolytic uremic syndrome.

Enterohemorrhagic Escherichia coli (EHEC) cause hemorrhagic colitis and hemolytic uremic syndrome (HUS), make potent cytotoxins (Verotoxins [VT] or Shiga-like toxins), and possess a plasmid (approximately 60 megadaltons) that encodes a new fimbrial antigen and promotes attachment to epithelial cells. We evaluated the use of a DNA probe, prepared from a 3.4-kilobase segment of the EHEC plasmid, to detect EHEC. The probe hybridized with 106 (99%) of 107 O157:H7 and 34 (77%) of 44 O26:H11, VT-positive strains from patients with colitis, HUS, and diarrheal disease and hybridized with 21 (81%) of 26 VT-positive E. coli of serotypes other than O157:H7 or O26:H11 from patients with hemorrhagic colitis and HUS. We examined 601 other strains, including 18 serotype O26 isolates of H types other than H11, 306 enteropathogenic E. coli, 60 enteroinvasive E. coli, 119 enterotoxigenic E. coli, and 20 isolates from the urinary tract and 77 isolates from the normal intestinal flora; only one (O127:H-) was positive (specificity, 99.8%). Serotype O26:H11, previously considered a classic enteropathogenic E. coli serotype, is now shown to be EHEC.

Qualitative and quantitative determination of enterobacterial common antigen (ECA) with monoclonal antibodies: expression of ECA by two Actinobacillus species.

The presence and quantity of the enterobacterial common antigen (ECA) in several species belonging to the family Enterobacteriaceae as well as to other gram-negative families were determined by a solid-phase enzyme-linked immunosorbent assay system and Western blotting by using mouse monoclonal antibodies specific for ECA. Except for Erwinia chrysanthemi, previously known to be an exception, all species known or presumed to belong to Enterobacteriaceae produced ECA (89 of 90 species). Most species not belonging to Enterobacteriaceae did not produce ECA (25 of 28 species), with one already known (Plesiomonas shigelloides) and two hitherto unknown (Actinobacillus equuli and Actinobacillus suis) exceptions. Interestingly, all strains of P. shigelloides produced ECA, regardless of the presence of the Shigella sonnei cross-reacting O antigen. Quantitation of the amount of ECA in members of the family Enterobacteriaceae revealed a remarkable heterogeneity among genera and species as well as within one species. We conclude that the rapid, sensitive, and reliable determination of ECA is a useful aid in taxonomic classification and may help to characterize the relatedness of the family Enterobacteriaceae to other families. However, a quantitative analysis of ECA appears to be without value for these purposes.

Plasmid analysis of simultaneous nosocomial outbreaks of methicillin-resistant Staphylococcus aureus.

A large outbreak of infections caused by methicillin and aminoglycoside resistant Staphylococcus aureus provided the opportunity to evaluate mechanisms of resistance and compare the usefulness of typing systems. Between January 1979 and December 1980, 63 patients developed infections with S aureus resistant to multiple antibiotics, including methicillin and tobramycin. All isolates had an identical antibiogram and were phage type 47/54/75/77/83A. Beginning in January 1981, a superimposed outbreak caused by S aureus of the same phage type but with a resistance pattern now including gentamicin occurred. The two strains contained different aminoglycoside inactivating enzymes. The initial strain contained a single plasmid of 21.5 mDa molecular weight, whereas the subsequent strain which had acquired gentamicin resistance contained this plasmid plus a heavier one of 33 mDa. Plasmid analysis complements the analysis of antibiograms and phage types and aids in defining epidemiologic patterns of transmission.

Molecular epidemiology of bacterial infections: examples of methodology and of investigations of outbreaks.

Plasmid profile analysis, plasmid and chromosomal bacterial restriction endonuclease DNA analysis, and DNA hybridization are increasingly used in clinical microbiology and epidemiology. These techniques have been applied, singly and in combination, to investigations of outbreaks, including those of diarrheal diseases caused by pandemic Vibrio cholerae, enterotoxigenic and enterohemorrhagic Escherichia coli, Salmonella muenchen, and Salmonella typhimurium. The techniques have been critical in the unraveling of outbreaks of disease caused by nosocomial and community-acquired pathogens. Although there are limitations to these methods, their applications have important ramifications for basic science and public health.

Plasmid characterization of Salmonella typhimurium transmitted from animals to humans.

The transmission of pathogenic bacteria from animals to humans is widely studied because of its public health importance. In this study, we show the transmission of Salmonella typhimurium from cattle which had received no growth-promoting antibiotics to humans who had direct contact with the ill animals. On one cattle farm, the veterinarian attending the sick animals became ill, and two other individuals living on the farm later developed salmonellosis. The strains isolated from both humans and animals at one farm were identical as to antibiotic susceptibility and phage type, and they were specifically traced by the presence of a common 24-megadalton plasmid. Restriction enzyme digests of this plasmid from both human and animal strains were identical. At another farm, tetracycline-resistant S. typhimurium strains possessing a different profile (eight plasmids) were isolated from both animals and humans. The tetracycline-resistant clone was also isolated from animals at a third farm, but with animals and humans having no known contact with those of the other two farms.

A multistate outbreak of gastrointestinal illness caused by enterotoxigenic Escherichia coli in imported semisoft cheese.

In September 1983, three clusters of gastrointestinal illness with similar symptoms affected 45 persons in Washington, D.C., after office parties. The illness lasted a mean of 4.4 days and was characterized by watery diarrhea (91%), abdominal cramps (80%), headache (38%), nausea (38%), and subjective fever (20%). Illness was strongly associated with having eaten imported French Brie cheese one to six days before onset of illness (P less than .0001 by Fisher's two-tailed exact test). After publicity about these outbreaks, additional cheese-associated cases were identified over an eight-week period in Illinois, Wisconsin, Georgia, and Colorado. Stool specimens from ill persons in four states yielded Escherichia coli serotype O27:H20. These organisms produced heat-stable enterotoxin and had similar plasmid profiles. When commercially distributed foods are contaminated, enterotoxigenic E. coli can cause widespread disease even in a developed country, and the disease can easily escape correct diagnosis.

Genotypic approaches to the diagnosis of bacterial infections: plasmid analyses and gene probes.

Practical genetic approaches have been helpful in the diagnosis, epidemiology, and taxonomy of bacterial pathogens encountered in our laboratory at the Centers for Disease Control. There are many examples in which plasmid profiles have been used to define epidemic strains of enteric bacteria, staphylococci, pseudomonads, vibrios, and other pathogenic bacteria. Current methodologies should allow the microbiology laboratory to use plasmid profiles routinely and to identify plasmids associated with bacterial pathogenesis. Simplified DNA-DNA hybridization procedures have been used in our laboratory to survey or "probe' thousands of Escherichia coli colonies for the presence of enterotoxin genes, eliminating traditional tissue culture or animal assays. Research scientists continue to develop gene probes for a number of bacterial toxins and hemolysins and for the identification of pathogens such as legionellae and salmonellae. These and other probes as well as hybridization "kits' may be commercially available to diagnostic laboratories within the next few years.