First Experimental Evidence for the Presence of Potentially Virulent Klebsiella oxytoca in 14 Species of Commonly Consumed Aquatic Animals, and Phenotyping and Genotyping of K. oxytoca Isolates

Klebsiella oxytoca is a recently emerging pathogen that can cause necrotizing enterocolitis, hemorrhagic colitis, sepsis-associated purpura fulminans, and infective endocarditis in humans. The bacterium is ubiquitous in water and soil environments. Nevertheless, current literature on K. oxytoca in aquatic products is rare. In this study, we surveyed K. oxytoca contamination in 41 species of consumable aquatic animals sold in July, August, and September of 2018 and 2019 in Shanghai, China, 40 of which had no history of carrying this bacterium. K. oxytoca was for the first time isolated from 14 species with high abundance in benthic animals. None of the K. oxytoca isolates (n = 125) harbored toxin genes mviM, tisB, and yqgB. However, a high occurrence of virulence-associated genes was observed, including brkB (73.6%), cdcB (66.4%), pduV (64.8%), and virk (63.2%). Resistance to sulphamethoxazole-trimethoprim (56.0%) was the most predominant among the isolates, followed by chloramphenicol (6.4%), tetracycline (5.6%), and kanamycin (3.2%). Approximately 8.0% of the isolates displayed multidrug resistant phenotypes. Meanwhile, high percentages of the isolates tolerated the heavy metals Cu2+ (84.8%), Pb2+ (80.8%), Cr3+ (66.4%), Zn2+ (66.4%), and Hg2+ (49.6%). Different virulence and resistance profiles were observed among K. oxytoca isolates in 3 types and 14 species of aquatic animals. The ERIC-PCR-based genome fingerprinting of the 125 K. oxytoca isolates revealed 108 ERIC genotypes with 79 singletons, which demonstrated the genetic diversity of the isolates. The results of this study fill gaps for policy and research in the risk assessment of K. oxytoca in consumable aquatic animals.


Introduction
Klebsiella oxytoca is a Gram-negative bacterium that ubiquitously resides in water and soil environments [1,2]. The bacterium is also found in the nasopharynx and intestine of healthy individuals in humans [3]. K. oxytoca was originally isolated from a yogurt specimen in 1886 and named Bacillus oxytoca. In 1963, this organism was taxonomically classified as a member of the genus Klebsiella [4]. Recently, it has been reported that K. oxytoca is the causative agent of necrotizing enterocolitis [5], hemorrhagic colitis [6], sepsis-associated purpura fulminans [7], and infective endocarditis in humans [8], arguing that K. oxytoca is an emerging pathogen [1].
Outbreak of K. oxytoca in hospitals involved strains with extended-spectrum betalactamases and carbapenemases, which consequently lead to therapeutic problems [9]. the 125 K. oxytoca strains were detected positive in the capsular staining but negative in Gram's staining and the dynamic tests. The results were confirmed by 16S rRNA gene sequencing and analysis. Among the 125 K. oxytoca isolates, approximately 88.0% (n = 110), 11.2% (n = 14), and 0.80% (n = 1) were recovered from the mollusk, fish, and crustacean samples, respectively (Table S1). Approximately 77.6% (n = 97) and 22.4% (n = 28) of the isolates were derived from 11 and 3 species of seawater and freshwater animal samples, respectively. Remarkably, most (96.8%, n = 121) of the K. oxytoca isolates were recovered from benthic aquatic animals.
The 125 K. oxytoca isolates were recovered from aquatic animal products originating from four provinces and one city located along the East China Sea, one of the major fishing grounds along China's coast, which encompasses the Fujian, Jiangsu, Shandong, and Zhejiang provinces, as well as Shanghai City in China. The highest abundance of K. oxytoca isolates were observed in the samples originating from Zhejiang Province (32.8%, 41 of the 125 isolates), followed by Fujian Province (25.6%, 32/125), Shandong Province (20.8%, 26/125), and Shanghai City (19.2, 24/125). Only a few isolates (1.6%, 2/125) originated from Jiangsu Province. Approximately 91.2% of the 125 K. oxytoca were isolated from the samples collected in the Luchao Port Aquatic Market, and 8.8% from the Jiangyang Aquatic Market in Shanghai, China.
The K. oxytoca isolates recovered from the three types of aquatic products had different virulence-associated gene profiles. Notably, high percentages of the brkB, cdcB, pduV genes were observed in the K. oxytoca isolates recovered from the mollusk (72.7-64.5%) and fish samples (78.6-57.1%). Meanwhile, relatively higher detection frequencies of the relE, and symE genes were found in the mollusks (44.5%, and 40.9%) than in the fish (28.6%, and 7.1%). The vagC and virk genes were only present in the isolates recovered from these two types of aquatic products, but with higher incidence from the mollusks (23.6%, 70.0%) than in the fish (21.4%, 14.3%). Additionally, the isolate (K. o-P. clarkii 8-1-12-7) from the crustacean (P. clarkia) carried the brkB, cdcB, relE, pduV, and symE genes (Figure not shown).
Additionally, different antibiotic resistance profiles were also observed among the K. The K. oxytoca isolates recovered from the 3 types of aquatic products had different antibiotic resistance profiles (Figure not shown). Higher rates of resistance to SXT were observed in the isolates recovered from the mollusks (60.9%, 67/110) than those from the fish (21.4%, 3/14) and crustacean (0%, n = 1). The CHL, TET, KAN, and GEN resistance were solely present in the isolates from the mollusks (7.3%, 6.4%, 3.6%, 0.9%), whereas CIP and NOR resistance were only found in the isolates derived from the fish (7.1%, and 7.1%) (Figure not shown).
The multiple antimicrobial resistance index (MARI) values of the 125 K. oxytoca isolates ranged from 0.44 to 0.00, which indicated varying degrees of exposure to the nine antibiotics evaluated. The mean MARI values of the isolates originating from the mollusks, fish, and crustacean were 0.21, 0.15, and 0, respectively. Among the 14 species of aquatic animals, the maximum MARI value was found from the isolates recovered from N. cumingi Crosse (0.44), and H. rubra (0.44), followed by C. auratus (Crucian) (0.33), M.
The multiple antimicrobial resistance index (MARI) values of the 125 K. oxytoca isolates ranged from 0.44 to 0.00, which indicated varying degrees of exposure to the nine antibiotics evaluated. The mean MARI values of the isolates originating from the mollusks, fish, and crustacean were 0.21, 0.15, and 0, respectively. Among the 14 species of aquatic animals, the maximum MARI value was found from the isolates recovered from N. cumingi Crosse
K. oxytoca isolates recovered from the three types of aquatic products had different heavy metal tolerance profiles. The majority of the isolates from all three types of aquatic products were tolerant to Cu 2+ (100% to 84.5%), Hg 2+ (100% to 48.2%), Pb 2+ (100% to 79.1%) and Zn 2+ (100% to 61.8%). Tolerance to Cr 3+ was prevalent among the isolates from the fish (100%) and mollusks (62.7%), whereas lower percentages of Cd 2+ tolerance were found in the isolates from the mollusks (10.0%) and fish (7.1%). Tolerance to Mn 2+ (12.7%) and Ni 2+ (0.9%) were solely observed in the isolate of the mollusks (Figure not shown).

Genetic Diversity of the K. oxytoca Isolates
The enterobacterial repetitive intergenic consensus-PCR (ERIC-PCR) was used to analyze the genetic diversity of the 125 K. oxytoca isolates recovered from the 14 species of aquatic animals. The obtained genome fingerprinting profiles comprised various numbers of DNA bands mainly ranging from 100 to 1000 bp (Figure 5a,b). Based on the fingerprinting profiles, all the isolates were classified into 108 different ERIC-genotypes, 73.2% of which were assigned as singletons (n = 79). Approximately 36.7% (n = 29), 30.4% (n = 24), and 25.3% (n = 20) of these singletons were derived from the mollusks N. cumingi Crosse,

Genetic Diversity of the K. oxytoca Isolates
The enterobacterial repetitive intergenic consensus-PCR (ERIC-PCR) was used to analyze the genetic diversity of the 125 K. oxytoca isolates recovered from the 14 species of aquatic animals. The obtained genome fingerprinting profiles comprised various numbers of DNA bands mainly ranging from 100 to 1000 bp (Figure 5a,b). Based on the fingerprinting profiles, all the isolates were classified into 108 different ERIC-genotypes, 73.2% of which were assigned as singletons (n = 79). Approximately 36.7% (n = 29), 30.4% (n = 24), and 25.3% (n = 20) of these singletons were derived from the mollusks N. cumingi Crosse, T. granosa, and A. woodiana, respectively. The UPGMA algorithm grouped all the 108 ERIC genotypes into 14 distinct clusters (clusters I to XIV) at a 32.0% similarity cut-off level ( Figure 5). Approximately 14.4% (n = 18) of the K. oxytoca isolates were classified into the largest cluster, II, followed by 11.2% (n = 14), 9.6% (n = 12), and 8.0% (n = 10) into clusters VIII, III, and I, respectively. The remaining isolates (56.8%) fell into clusters III-V, VII, and IX-XIV (7.2% to 0.8%). Most isolates had a similarity coefficient of 30.0-85.0%, and a Simpson's diversity index of 0.8485. These results demonstrate the considerable genetic diversity of the 125 K. oxytoca isolates recovered from the 14 species of aquatic animals.

Comparison of the MDR and Heavy Metal Tolerance
To get insights into co-selection between antibiotics and heavy metals in K. oxytoca, we further analyzed the 10 isolates with MDR phenotypes by phylogenetic analysis (Figure 6). The resulting data revealed the genetic diversity of the MDR K. oxytoca isolates with a Simpson's index of 0.8556. Moreover, the MDR isolates belonging to 8 ERIC-genotypes were classified into 5 distinct clusters (clusters A-E) ( Figure 6).

Comparison of the MDR and Heavy Metal Tolerance
To get insights into co-selection between antibiotics and heavy metals in K. oxytoca, we further analyzed the 10 isolates with MDR phenotypes by phylogenetic analysis (Figure 6). The resulting data revealed the genetic diversity of the MDR K. oxytoca isolates with a Simpson's index of 0.8556. Moreover, the MDR isolates belonging to 8 ERIC-genotypes were classified into 5 distinct clusters (clusters A-E) ( Figure 6).

Discussion
K. oxytoca is the second most common Klebsiella species after K. pneumoniae [5]. Nevertheless, systematic research on K. oxytoca is still in its infancy, and current literature on K. oxytoca in aquatic products is rare [1,20]. Recently, Håkonsholm et al. reported 40 K. oxytoca strains isolated from M. edulis and one from C. gigas [2]. In this study, we surveyed The cytotoxin-producing K. oxytoca has recently been identified as a new candidate etiologic agent in the pathogenesis of necrotizing enterocolitis [5]. In fact, information in genes associated with the virulence of K. oxytoca still remains to be discovered. Previous studies have indicated that the genes encoding adhesins, siderophores, and invasins of the genus Klebsiella were associated with deleterious traits [21]. In this study, virulencerelated genes (brkB, cdcB, pduV, relE, symE, vagC, and virK) in the 125 K. oxytoca isolates recovered from 14 species of aquatic animals were detected simultaneously. High incidence of the brkB (73.6%), cdcB (66.4%), pduV (64.8%), and virk (63.2%) genes was observed. The brkB gene encodes the YihY/virulence factor BrkB family protein, while cdcB, pduV, and virK encode a cytotoxin, a type I toxin-antitoxin system hok family toxin, and the virulence factor VirK, respectively [22,23]. The RelE toxin in Escherichia coli was a global inhibitor of translation [24]. SymE was toxic in the inhibition of protein synthesis and RNA degradation [25]. The virulence factor VagC was involved in the bacteriocin secretion system and the type II toxin-antitoxin system in Turkish Salmonella serovar Infantis isolates [26]. In this study, the virulence-associated genes mviM, tisB, and yqgB were absent from the K. oxytoca isolates. As a virulence gene, mviM was associated with the exercise, biofilm formation and antimicrobial resistance of Cronobacter sakazakii [27]. TisB is a component of the TisB/IstR-1 toxin-antitoxin system in E. coli [28]. The adaptive factor YqgB in Bacillus thuringiensis facilitated the bacterial colonization of the host [29]. In this study, the K. oxytoca isolates recovered from the 3 types and 14 species of aquatic animals harbored different virulence-associated gene profiles. Remarkably, the isolates originating from A. woodiana, N. cumingi crosse, S. subcrenata, and T. granosa harbored the maximum number (n = 7) of the virulence-associated genes tested, which suggests a health risk from the potentially virulent K. oxytoca in these consumable aquatic animals.

Discussion
K. oxytoca is the second most common Klebsiella species after K. pneumoniae [5]. Nevertheless, systematic research on K. oxytoca is still in its infancy, and current literature on K. oxytoca in aquatic products is rare [1,20]. Recently, Håkonsholm et al. reported 40 K. oxytoca strains isolated from M. edulis and one from C. gigas [2]. In this study, we surveyed K. oxytoca contamination in 41 species of commonly consumed aquatic animal products sold in Shanghai in July, August, and September of 2018 and 2019. K. oxytoca was for the first time isolated from 14 species of aquatic animals, including 10 species of mollusks: A. woodiana, B. areolata, C. cahayensis, H. rubra, M. antiquata, M. edulis, N. cumingi Crosse, S. subcrenata, S. constricta and T. granosa; 3 species of fish: B. rock cod, C. auratus (Crucian), and C. auratus (Ditrema temmincki Bleeker); and one species of crustacean: P. clarkii. Our data also provided the first experimental evidence for high detection frequencies of K. oxytoca in benthic aquatic animals, such as N. cumingi crosse, T. granosa, and A. woodiana. These results suggested a potential health risk of the bacterial transmission to communities through the aquatic animals.
The cytotoxin-producing K. oxytoca has recently been identified as a new candidate etiologic agent in the pathogenesis of necrotizing enterocolitis [5]. In fact, information in genes associated with the virulence of K. oxytoca still remains to be discovered. Previous studies have indicated that the genes encoding adhesins, siderophores, and invasins of the genus Klebsiella were associated with deleterious traits [21]. In this study, virulence-related genes (brkB, cdcB, pduV, relE, symE, vagC, and virK) in the 125 K. oxytoca isolates recovered from 14 species of aquatic animals were detected simultaneously. High incidence of the brkB (73.6%), cdcB (66.4%), pduV (64.8%), and virk (63.2%) genes was observed. The brkB gene encodes the YihY/virulence factor BrkB family protein, while cdcB, pduV, and virK encode a cytotoxin, a type I toxin-antitoxin system hok family toxin, and the virulence factor VirK, respectively [22,23]. The RelE toxin in Escherichia coli was a global inhibitor of translation [24]. SymE was toxic in the inhibition of protein synthesis and RNA degradation [25]. The virulence factor VagC was involved in the bacteriocin secretion system and the type II toxin-antitoxin system in Turkish Salmonella serovar Infantis isolates [26]. In this study, the virulence-associated genes mviM, tisB, and yqgB were absent from the K. oxytoca isolates. As a virulence gene, mviM was associated with the exercise, biofilm formation and antimicrobial resistance of Cronobacter sakazakii [27]. TisB is a component of the TisB/IstR-1 toxin-antitoxin system in E. coli [28]. The adaptive factor YqgB in Bacillus thuringiensis facilitated the bacterial colonization of the host [29]. In this study, the K. oxytoca isolates recovered from the 3 types and 14 species of aquatic animals harbored different virulence-associated gene profiles. Remarkably, the isolates originating from A. woodiana, N. cumingi crosse, S. subcrenata, and T. granosa harbored the maximum number (n = 7) of the virulence-associated genes tested, which suggests a health risk from the potentially virulent K. oxytoca in these consumable aquatic animals.
The propagation and spread of resistant pathogenic bacteria pose serious threats to the public heath for humans and animals [30,31]. It is estimated that antibiotic resistance may lead to 10 million deaths per year by 2050 [31]. Misuse of antimicrobial agents is the main cause of antibiotic resistance in pathogenic bacteria, particularly in developing nations [32][33][34]. In this study, our results indicate that SXT resistance was the most predominant (56.0%) among the 125 K. oxytoca isolates of aquatic animal origins. SXT-resistant K. oxytoca isolates in humans has also been reported. For example, Maharjan et al. recently reported that 9.4% of K. oxytoca strains (n = 48) isolated from stool specimens of healthy adult volunteers (n = 510) in Kathmandu in Nepal showed the highest resistance toward SXT (45.8%) [35]. In this study, incidences of intermediate susceptibility to KAN (40.8%), and CIP (12.0%) were observed, which suggests a potential resistance trend of K. oxytoca in aquaculture environments.
MARI is commonly used to determine the health risk associated with antibiotic resistance [17]. In this study, the mean MARI values for K. oxytoca isolates derived from the mollusks, fish, and crustacean samples were 0.21, 0.15 and 0, respectively. Among the 14 species of aquatic animals, the maximum MARI value was derived from the isolates in H. rubra  [18], which provided direct evidence for the antibiotic-resistant phenotypes of K. oxytoca isolates observed in this study.
A high level of the bioaccumulation of toxic heavy metals (e.g., Cd 2+ , Cr 3+ , Hg 2+ , Ni 2+ , and Pb 2+ ) through the food chain is a grave threat to human health, due to their non-degradable nature [36]. Numerous studies have reported heavy metal residues in various aquatic environments and species of food animals (shellfish, fish, crustaceans, and crabs) sampled worldwide, particularly in developing nations [13][14][15]18,37]. In the recent surveys by our research group [17,38,39], the heavy metal-tolerant waterborne pathogen Vibrio cholerae has been discovered in many species of aquatic animals. In this study, the results indicated that the K. oxytoca isolates originated from the 3 types and 14 species of aquatic animals had different heavy metal tolerance profiles. Tolerance to Cu 2+ and Pb 2+ was the most prevalent among the isolates (84.8% and 80.8%). Cu is an essential element for various biological functions in many organisms, and it is also a cofactor for hemocyanin in aquatic arthropods and mollusks. However, excessive levels of Cu can be toxic or disrupting metabolic processes [40]. In this study, our data also revealed a higher detection frequency of Cr 3+ tolerance in the K. oxytoca isolates from the fish samples than that from the mollusks. Moreover, the isolates recovered from S. constricta were tolerant to all 8 heavy metals, followed by H. rubra, and T. granosa (7 heavy metals); C. auratus (Crucian), and S. subcrenata (6 heavy metals); B. areolata, B. rock cod, C. auratus (Ditrema temmincki Bleeker), and N. cumingi Crosse (5 heavy metals). Tolerance to Ni 2+ was only detected in the isolates from S. constricta. These data suggest serious heavy metal pollution likely occurred in the aquaculture environments, consistent with previous reports [18,38,39]. Recently, Ni et al. reported that the heavy metals Cu, Hg, Pb, and Cd were observed in the 41 species of aquatic animals with PSRs of 100%, 100%, 77.4%, and 34.0%, respectively, none of which exceeded their MRLs [18], which provided direct evidence for the high incidence of K. oxytoca tolerance to heavy metals in this study. Additional attention should be paid to the potential health risk of heavy metal pollution in consumable aquatic animals sold in Shanghai, China.
In this study, ten K. oxytoca isolates had MDR phenotypes, nine of which also exhibited tolerance to two or more than two heavy metals. For example, the K. oxytoca isolate K. o-H. rubra 8-2-2-11, showing resistance to the four antibiotics CHL, SXT, KAN, and TET, also tolerated the 6 heavy metals Cr 3+ , Cu 2+ , Hg 2+ , Mn 2+ , Pb 2+ , and Zn 2+ . These results suggested that heavy metal pollution likely co-selected for antibiotic resistance in K. oxytoca, and vice versa. In our previous research, antibiotic resistance correlating positively with heavy metal resistance was also observed in the V. cholerae isolates recovered from the 41 species of aquatic animals [18]. Cross-resistance due to co-selection can be inferred as the most likely mechanism of the rising antibiotic-resistant pathogens [41].
ATCC43165 (Guangdong Culture Collection Center, Guangzhou, China) was used as a positive control strain in this study.

Statistical Analysis
Data analysis was performed using the SPSS statistical analysis software, version 17.0 (SPSS Inc., Chicago, IL, USA). The MARI of an isolate was defined as a/b, where a represents the number of antibiotics to which the isolate was resistant, while b represents the number of antibiotics for which the isolate was examined [17,38,39,45].