Multistate reptile‐ and amphibian‐associated salmonellosis outbreaks in humans, United States, 2009–2018

Abstract Non‐typhoidal Salmonella cause an estimated 1.4 million human illnesses, 26,000 hospitalizations and 400 deaths annually in the United States. Approximately 11% of these infections are attributed to animal contact. Reptiles and amphibians are known sources of salmonellosis; young children (aged <5 years) are disproportionately affected by reptile‐ and amphibian‐associated salmonellosis (RAAS) outbreaks. We describe multistate RAAS outbreaks to characterize illnesses and inform prevention efforts. RAAS outbreaks were defined as ≥2 culture‐confirmed human Salmonella infections with similar pulsed‐field gel electrophoresis patterns and epidemiologic, laboratory or traceback evidence linking them to a common reptile/amphibian exposure. Data sources included the Animal Contact Outbreak Surveillance System; CDC Outbreak Response and Prevention Branch's outbreak management database; PulseNet, the national molecular subtyping network for foodborne disease surveillance in the United States; and the National Antimicrobial Resistance Monitoring System. Twenty‐six RAAS outbreaks were reported during 2009–2018, resulting in 1465 illnesses and 306 hospitalizations. The outbreaks were associated with turtles (19), lizards (5), snakes (1) and frogs (1). Sixteen (61.5%) outbreaks were linked to small turtles (<4 inches), resulting in 914 illnesses. Forty‐nine percent of outbreak‐associated patients were aged <5 years. Of 362 patients/caregivers interviewed, 111 (30.7%) were aware that reptiles/amphibians can carry Salmonella. Among 267 patient isolates with antimicrobial susceptibility information, 20 (7.5%) were non‐susceptible to ≥1 antibiotic used to treat human salmonellosis. RAAS outbreaks result in considerable morbidity, particularly among young children. Illnesses linked to small turtles are preventable through education, targeted outreach to caregivers and paediatricians, and when appropriate, enforcement. Historically, individual states and jurisdictions have enforced existing or promulgated new authorities to address outbreaks. Preventing future RAAS outbreaks requires addressing challenges related to the illegal sale/distribution of small turtles; and for legal reptile sales, providing information on RAAS risk to consumers at point of sale to support informed pet ownership decisions.


| INTRODUC TI ON
Non-typhoidal Salmonella cause an estimated 1.4 million human illnesses, 26,000 hospitalizations and 400 deaths annually in the United States (Collier et al., 2021). Salmonella infection typically causes a self-limiting gastroenteritis (e.g. diarrhoea, fever and abdominal pain); however, young children (<5 years of age), persons ≥65 years of age and immunocompromised individuals are at greater risk for serious complications, including septicaemia, meningitis and death (Wen et al., 2017).
Historically, Salmonella infections linked to pet turtles were so common that a federal ban on the sale of small turtles (<4 inches) was implemented in 1975 (Food and Drug Administration, 2019).
The prevalence of Salmonella carriage can be >90% in reptiles and amphibians, and studies have documented Salmonella detection from pet reptiles and surfaces in reptile-owning households (Back et al., 2016;Clancy et al., 2016;Lowther et al., 2011;Nakadai et al., 2005). Approximately 2.9% of U.S. households owned reptiles in 2016, representing a 17% increase in reptile ownership since 2011 (American Veterinary Medical Association, 2018).
Although most non-typhoidal Salmonella infections do not warrant antibiotic treatment, administration of antibiotics is recommended to treat severe infections such as bacteraemia and meningitis and may benefit patients at increased risk for invasive infection (Pegues & Miller, 2015). Antimicrobial resistance has been identified among Salmonella isolated from pet reptiles and amphibians and from people linked to RAAS outbreaks, raising concern regarding the potential for these pets to serve as sources of resistant bacteria resulting in human infection (Centers for Disease Control and Prevention, 2014;Centers for Disease Control and Prevention, 2016a;Chen et al., 2010;Guerra et al., 2010).
We aimed to describe the epidemiology of multistate RAAS outbreaks in the United States from 2009 to 2018. We quantified the number of outbreaks and associated cases, and described the trends, patient demographics, types of human-animal interactions most likely to lead to infection, and antimicrobial susceptibility of strains causing RAAS outbreaks.

| Outbreak detection and case identification
We defined multistate RAAS outbreaks as outbreaks with ≥2 culture-confirmed human Salmonella infections with exposures reported from ≥2 states or territories with similar pulsed-field gel electrophoresis (PFGE) patterns and with a combination of epidemiologic, laboratory and traceback evidence linking them to contact with a common reptile or amphibian species during [2009][2010][2011][2012][2013][2014][2015][2016][2017][2018] (Centers for Disease Control and Prevention, 2020). We defined reptile or amphibian contact as contact with the animal, its bodily fluids, food or contact with the environment where the animal resides. For each outbreak, we defined a case as an illness in a person infected with a laboratory-confirmed Salmonella isolate matching the outbreak PFGE pattern and occurring while the outbreak investigation was ongoing (date of first patient illness onset to date outbreak investigation was declared over by CDC). Data sources included the Animal Contact Outbreak Surveillance System (ACOSS); CDC Outbreak Response and Prevention Branch's outbreak management database; PulseNet, the national laboratory network for molecular subtyping of Salmonella and other enteric pathogens; and the illegal sale/distribution of small turtles; and for legal reptile sales, providing information on RAAS risk to consumers at point of sale to support informed pet ownership decisions.

K E Y W O R D S
amphibians, antimicrobial resistance, microbial, reptiles, salmonella, zoonotic diseases Impacts • Reptile and amphibian contact remains an important source of salmonellosis in the United States, with young children disproportionately affected by reptile-and amphibian-associated salmonellosis (RAAS) outbreaks.
• Most multistate RAAS outbreaks and outbreakassociated illnesses were linked to small turtles.
Occurrence of these preventable outbreaks highlights challenges with enforcing regulations prohibiting their sale/distribution as pets.
• Non-susceptibility to clinically important antibiotics in Salmonella causing RAAS outbreaks is a public health concern because it can make severe infections harder to treat.
• To prevent future RAAS outbreaks, information on RAAS risks should be readily available to consumers at point of sale to support informed pet ownership decisions.
National Antimicrobial Resistance Monitoring System (NARMS). We reviewed data on reptile and amphibian exposures collected from patients (or their parents/guardians) during outbreak investigations.

| Outbreak and patient characteristics
We examined outbreak characteristics and trends over time, including number of outbreaks per year, type of reptile or amphibian implicated, outbreak size and duration, specific Salmonella serotype(s), and isolates from specimens collected from animals or their environment (e.g. habitat, water, bedding). We classified reptile or amphibian type into five categories: turtles ≥4 inches, turtles <4 inches, lizards (e.g. bearded dragons, geckos), snakes and frogs. Isolates that had serotypes that were undetermined, pending or did not match the predominant outbreak strain were individually reviewed and determined to be consistent with the predominant outbreak serotypes based on PFGE pattern. Therefore, for the purposes of this analysis, these isolates were categorized as the predominant serotype if their PFGE pattern was indistinguishable from the outbreak PFGE pattern.
We examined patient characteristics including demographic information (e.g. age and sex), health outcomes (e.g. hospitalizations and deaths) and isolate source (e.g. stool, urine or blood). We defined infants as children <1 year of age. We considered the identification of an isolate from a blood sample as an indicator for a bloodstream infection. When patient characteristics were unknown, the data were considered missing for this analysis. Reported deaths were examined, and those not attributable to salmonellosis were excluded from this analysis. We compared outbreak and patient characteristics of small turtle outbreaks to all other RAAS outbreaks to examine differences by category of reptile or amphibian implicated. We analysed common pet ownership and handling practices reported among patients.
We defined isolates as non-susceptible if they had a minimum inhibitory concentration (MIC) in the resistant or intermediate range by AST or a resistance mechanism (gene or mutation) identified by ResFinder and PointFinder screening of assembled WGS data; isolates with a fosA7 gene and no other resistance mechanisms were considered susceptible. We defined clinically important antibiotics as those used to treat salmonellosis in people, including ampicillin, azithromycin, ceftriaxone, ciprofloxacin and trimethoprim-sulfamethoxazole; isolates with non-susceptibility (phenotypic or genotypic) to any of these agents were considered to have clinically important resistance

| Statistical methods
We determined the median and range for outbreak size and duration and calculated the mean number of outbreaks that occurred during two time periods: 2009-2013 and 2014-2018. We calculated frequencies for outbreak variables (category of reptile or amphibian implicated, Salmonella serotype) and patient variables (age, sex, hospitalizations, deaths, isolate source and pet ownership and handling practices). We described hospitalization and isolate source by patient age. We compared outbreak size, duration, median age and proportion of patients hospitalized by outbreak type (small turtle vs. all other RAAS outbreaks) using the Wilcoxon rank-sum test and the Chi-square test, respectively and resulting two-sided p-values.
More than 60% of all RAAS outbreak-associated illnesses were linked to small turtles (Table 1). Overall, 69.2% (608/878) of patients in small turtle outbreaks were <18 years of age. The median patient age (5 years) and age range (<1-100 years) did not differ substantially between small turtle outbreaks and all other RAAS outbreaks (p = .083). Among all patients in RAAS outbreaks with known hospitalization status (992/1465; 67.7%), the proportion of patients hospitalized across all small turtle outbreaks was 31.1% compared to 30.4% across all other RAAS outbreaks (p = .846).

| Antimicrobial susceptibility
Antibiotic susceptibility information was available for 267 (18.2%) patient isolates from 24 (92.3%) outbreaks; 84 isolates were screened for antibiotic non-susceptibility by phenotypic AST, 138 by WGS and 45 by both methods (Figure 1, Table 3). Twenty-one (7.9%) patient isolates from nine outbreaks were non-susceptible to at least one antibiotic tested. Among these 21 patient isolates, 20 (95.2%) were non-susceptible to one or more clinically important antibiotics; of these, eight isolates were non-susceptible to two clinically important antibiotics and one was non-susceptible to three clinically important antibiotics. Additionally, antibiotic non-susceptibility information was available for 10.9% (39/359) of animal and environmental (e.g. habitat, water and bedding) isolates from nine outbreaks. One (2.6%) isolate, an environmental isolate from the snake outbreak, was nonsusceptible to a clinically important antibiotic. Various resistance mechanisms that confer non-susceptibility to fluoroquinolones, βlactam antibiotics and trimethoprim-sulfamethoxazole were identified in isolates with WGS data ( Table 4).
Patient isolates were most often non-susceptible to the fluoroquinolone ciprofloxacin. Thirteen (4.9%) patient isolates and one (2.6%) environmental isolate had a quinolone-resistance mechanism ( Phenotypic or genotypic non-susceptibility to β-lactam antibiotics was noted among eight patient isolates ( Table 4). Four patient isolates from three outbreaks (turtle <4 inches, bearded dragon and frog) were non-susceptible to ceftriaxone, ampicillin, cefoxitin and amoxicillin-clavulanic acid; of these, three isolates were sequenced and found to have the AmpC β-lactamase gene bla CMY-2 . Four additional patient isolates from two small turtle outbreaks were non-susceptible to ampicillin, including one isolate, which was also non-susceptible to amoxicillin-clavulanic acid. Three of these isolates had bla TEM genes.

No. (%) of nonsusceptible isolates
Five patient isolates from one bearded dragon outbreak were non-susceptible to trimethoprim-sulfamethoxazole; all five isolates were sequenced and found to have dfrA12, sul1 and sul3 genes.
Azithromycin non-susceptibility was not found among patient isolates. Non-susceptibility to other antimicrobials (e.g. aminoglycosides, phenicols and tetracyclines) was noted in some isolates (Table 4).

| DISCUSS ION
Reptile-and amphibian-associated salmonellosis outbreaks resulted in considerable morbidity, particularly among young children. These outbreaks are a reminder that contact with reptiles and amphibians remains an important source of salmonellosis in the United States.
Consistent with historical outbreaks of Salmonella linked to animal contact, RAAS outbreaks in this analysis were longer in duration compared to foodborne Salmonella outbreaks (Marshall et al., 2020;Marus et al., 2019). Reptiles and amphibians can be colonized with multiple strains of Salmonella, appear healthy while colonized and shed Salmonella intermittently; people might be exposed long after obtaining their pet, which can result in prolonged outbreak durations (Goupil et al., 2012;Hoelzer et al., 2011;Whitten et al., 2015).
Compared to all patients with non-typhoidal Salmonella infections (Jones et al., 2008), patients in RAAS outbreaks also had higher proportions of bloodstream infections (8.8% vs. 5%) and hospitalizations (30.8% vs. 22%). These findings indicate that infections associated with RAAS outbreaks continue to result in more adverse outcomes compared to other non-typhoidal Salmonella infections and outbreaks (Marus et al., 2019). Previous studies have reported blood isolation to be more common in serotype Sandiego than Typhimurium; however, this has not been previously reported for serotype Agbeni, as we found in this analysis (Angelo et al., 2016;Jones et al., 2008). Serotype Agbeni is an uncommon serotype in humans that appears to be emerging in the United States (Centers for Disease Control and Prevention, 2018).  Mitchell & Shane, 2001). Current practices to reduce or eliminate Salmonella in reptiles include using non-antibiotic compounds to treat water and/or eggs (Mitchell et al., 2007). Further evaluation to better understand drivers of antimicrobial resistance in pet reptiles and amphibians is needed, including antimicrobial use practices in the animal industry and spread of resistant bacteria on turtle farms.
a Includes patient isolates with antimicrobial susceptibility testing (AST) data or resistance mechanism data from whole genome sequencing (WGS). AST was performed using broth microdilution. We used Clinical and Laboratory Standards Institute breakpoints (where available) or consensus breakpoints from the National Antimicrobial Resistance Monitoring System. Whole genome sequence data was collected for select isolates by state public health laboratories or CDC using PulseNet standard methods and quality requirements. Antimicrobial resistance genes and mutations were identified from assembled sequences based on the ResFinder (updated May 23, 2019) and PointFinder (updated April, 29 2019) databases. We defined isolates as non-susceptible if they had a minimum inhibitory concentration in the resistant or intermediate range by AST or a resistance mechanism by WGS. b Three isolates with a fosA7 gene and no other resistance mechanism were considered susceptible.
c Plasmid loss likely occurred in this isolate, which had no resistance mechanisms detected by WGS and was susceptible on retesting. Ten isolates with a fosA7 gene and no other resistance mechanism were considered susceptible.

TA B L E 4 (Continued)
ciprofloxacin and ceftriaxone. Non-susceptibility to ciprofloxacin among patient isolates is noteworthy, as studies suggest that even small increases in quinolone MICs can result in poorer clinical outcomes and facilitate the emergence of fluoroquinolone resistance (Crump et al., 2003;Humphries et al., 2012;Rodríguez-Martínez et al., 2016). We identified PMQR genes among patient isolates, which is consistent with a recently published US study that identified PMQR genes in Salmonella isolates from patients with exposure to reptiles and amphibians . This finding is notable as PMQR genes may be transferred horizontally to other bacteria, thereby spreading resistance, and they can also facilitate the development of higher-level quinolone resistance (Robicsek et al., 2006;Rodríguez-Martínez et al., 2016). Some patient isolates were nonsusceptible to ceftriaxone, which is frequently recommended for treatment of invasive Salmonella infections, particularly in children (American Academy of Pediatrics, 2018). to the illegal sale and distribution of small turtles might result in decreased reptile-associated morbidity, especially among young children.

ACK N OWLED G EM ENTS
We gratefully acknowledge the state and local public health and animal health officials for their epidemiologic and laboratory support during the outbreak investigations described in this report. We appreciate the assistance and coordination among CDC's Outbreak

Response and Prevention Branch, National Outbreak Reporting
System, PulseNet, Enteric Diseases Laboratory Branch, and National Antimicrobial Resistance Monitoring System.

CO N FLI C T O F I NTE R E S T
None to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

D I SCL A I M ER
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers