The temporal dynamics of antimicrobial-resistant Salmonella enterica and predominant serovars in China

ABSTRACT Salmonella enterica is one of the most common bacterial pathogens in humans and animals. Systematic studies on the trends and geographical distribution of antimicrobial-resistant Salmonella and dominant serovars have been well studied in European and American countries while not in China. Here, taking the One-Health strategy, we used >35 000 Salmonella enterica isolates to explore the temporal and spatial dynamics of dominant serovars in China. We found that Salmonella Typhimurium was the dominant serovar causing human infection in China, which was consistent with Australia but inconsistent with North American and European countries. The proportion of Salmonella serovars Typhimurium, London, Rissen, Corvallis, Meleagridis, Kentucky, and Goldcoast showed an increasing trend during 2006–2019. We randomly selected 1962 isolates for comparative genomics and antimicrobial resistance studies and found that the number of antibiotic resistance genes (ARGs) per isolate increased 1.84 and 2.69 times of human and non-human origins, respectively, spanning 14 years. The proportion of antimicrobial-resistant Salmonella isolates had an increasing trend during 2006–2019, especially beta-lactam, quinolone, tetracycline, and rifampicin resistance. Moreover, we found that higher diversity of sequence types (STs) in S. Typhimurium than in other serovars, ST34 from pig and ST19 from chicken origin, were mainly associated with isolates causing child and adult gastro-infection, respectively. Our results fill in the data gap on the trends of dominant serovars and antimicrobial resistance of Salmonella enterica in China. These data provide useful information for public health decision-makers prioritizing interventions for foodborne diseases and food safety.


INTRODUCTION
Salmonella enterica is a major cause of global foodborne diseases, and antimicrobial-resistant Salmonella enterica isolates pose a serious threat to public health [1].Salmonella enterica consists of eleven subspecies [2].Salmonella enterica subspecies enterica serovars Typhi (S.Typhi) and Paratyphi (S.Paratyphi) A, B, or C can cause typhoid and paratyphoid fevers in humans, collectively referred to as enteric fever [3], whereas other serovars are loosely described as non-typhoidal Salmonella (NTS).It was estimated that 14.3 million and 53.5 thousand cases of enteric fever and NTS invasive disease occurred in 2017, respectively, [4,5].Salmonella enterica serovar Typhimurium (S.Typhimurium) was the most prevalent serovar causing human infection in Africa [6] and Australia [7].The S. Enteritidis was the dominant serovar causing human infections in the United States [8] and European countries [9].Recently, the World Health Organization (WHO) reported a multi-country outbreak of multidrug-resistant monophasic S. Typhimurium (S.I 1,4, [5],12: i:-ST34) infection linked to chocolate products (https://www.who.int/emergencies/diseaseoutbreak-news/item/2022-DON369),suggesting food vehicles associated with the transmission of this microbe.Although some studies reported that S. Typhimurium, S. Enteritidis, and S. I 1,4, [5],12: i:-were the most common serovars found among patients with diarrhea in China [10,11], most studies are from a single site, are of limited sample size, or of limited source and time.
Our previous study showed that a total of 322 serovars have been detected in China from 1951 to 2007 [12].In recent years, some new serovars such as S. Telkebier [13], S. Uzaramo [14], and S. Changwanni [15] have been detected in China by our colleagues, and all of them were identified and reported for the first time in foreign countries.Although the prevalence and diversity of Salmonella enterica serovars from the five continents (Africa, the Americas, Asia, Europe and Oceania) have been well summarized in a previous meta-analysis [16], there is currently no systematic study of the epidemiology, temporal and spatial dynamics of dominant serovars of Salmonella in China.To fill this data gap, we have systematically analyzed the features of >35 000 S. enterica strains isolated from human and non-human origins enrolled in the Chinese Local Surveillance System for Salmonella database (also called Bacterium-learning Union) across 23 provinces or municipal cities in China over 14 years (from 2006 to 2019), which can support basic research and public health risk forecasts for S. enterica.
The emergence and spread of antimicrobial resistance (AMR) in Enterobacteriaceae pose a serious threat to human and animal health [17,18].Notably, fluoroquinolone-resistant Salmonella spp.and extended-spectrum-beta-lactamase (ESBL)producing Enterobacteriaceae have been listed by the WHO in 2017 as among the high priority pathogens posing a risk to human health and need for the research and development of new antibiotics [17].It is estimated that a continued rise in AMR by 2050 will lead to 10 million people dying every year and a reduction of 2-3.5% in Gross Domestic Product [19].
Recently, whole-genome sequencing (WGS) has provided an unparalleled, powerful tool for understanding population dynamics, genomic epidemiology, and the investigation of outbreak dynamics, as well as the rapid temporal and spatial evolution of AMR in bacterial pathogens [20][21][22][23].WGS-based predictions of AMR and serovar were highly consistent with the phenotypic determination results [24][25][26][27].Moreover, comparative genomics could be performed to investigate the virulence profiles and mobile genetic elements (MGEs) of these isolates and their relationship with ARGs [28].Therefore, to explore the trends and prevalence of AMR, as well as the distinct features of isolates of human and non-human origin, we sequenced and analyzed 1962 S. enterica isolates (involving 130 serovars) across 22 provinces or municipal cities in China.These analyses are intended to provide useful information for the maintenance of national biosecurity and food safety interventions, especially for monitoring and preventing the importation and spread of AMR and S. enterica in the population.

Geographical distribution of 35 382 S. enterica isolates in China
From 1 September 1982, to 28 September 2019, 35 382 serotyped Salmonella spp.isolates (164 serovars) were registered at the Bacterium-learning Union (Fig. 1).The large collections were derived from 23 provinces or municipal cities in China (Fig. 1A).Of the 35 382 isolates, 34 804 (98.37%) belong to NTS, 419 (1.18%) to S. Paratyphi, 159 (0.45%) to S. Typhi (Fig. S1).A detailed description of the isolates is summarized in Fig. 1B and C. S. Typhimurium, S. Enteritidis, and S. Derby were the three most dominant serovars, accounting for 26.62%, 23.65% and 4.69%, respectively, (Fig. 1B).However, S. Typhi (115/486) was the most common among isolates from bloodstream infection samples, followed by S. Paratyphi A (101/486) (Fig. 1B).Overall, S. Typhimurium (23.73%, 8397/35 382) was the most common serovar in both human and non-human origins, followed by S. Enteritidis, S. Derby, S. London and S. Thompson (Fig. S2A).Results from the isolated S. enterica strains' serotyping show that in South and South-West regions, S. Typhimurium was the major serovar; in East, North, and North-West regions, S. Enteritidis was the major serovar (Fig. S2B & C).Moreover, we observed key differences between the distribution of Salmonellosis in China and other countries and regions.The most common serovar causing human infection was similar to Australia while inconsistent with the USA and England.In contrast to these countries, 4 (Meleagridis, London, Corvallis, Rissen) of the 10 most common serovars identified in our database did not appear on their lists (Table S1).

Dominant subtype switch in S. enterica serovars during 2006-2019 in China
To better understand the temporal dynamics of S. enterica serovars in China, a detailed description of the top 30 NTS serovars is summarized in Fig. 2. Generally, the proportion of Salmonella  and North-East contributed the smallest (Fig. 1B1), which were collected between 1991 and 2019 throughout the year (Fig. 1B2), a higher number of cases was observed between May and September (Fig. 1B3), and the ratio of males to females in this study was 1 : 1 (Fig. 1B4).The age of subjects ranged from 1 day to 90 years and 32.41% were between 0-9 years old (Fig. 1B5).S. Typhimurium, S. Enteritidis, and S. Derby were the ) three most dominant serovars, accounting for 26.62%, 23.65% and 4.69%, respectively (Fig. 1B6); 97.76% of the 25 884 strains were isolated from gastrointestinal samples (Fig. 1B7).However, the main source of extra-intestinal was bloodstream infection, S. Typhi (115/486) was the most common among isolates from bloodstream infection samples, followed by S. Paratyphi A (101/486) (Fig. 1B8).C, Summary of S. enterica strains isolated from non-human origin samples.C1, Collection region.C2, Collection year (1982-2019).C3, The proportions of fifteen dominant Salmonella serovars in China.C4, Source.C5, Chicken.C6, Pig.C7, Duck.C8, Isolates of bloodstream infection origin.Freshwater fish.C9, Turtle.C10, Marine.Of the 9498 isolates of non-human origin (a total of 120 serovars), S. Typhimurium was the most dominant serovar (Fig. 1C3), 87.54% (8315/9498) were isolated from food (Fig. 1C4) of animal origin (mainly including chicken, pig, duck, freshwater fishes, turtles, marine and cattle), and the distribution of serovars varied among different sources (Fig. 1C5-C10).
serovars Typhimurium, London, Rissen, Corvallis, Meleagridis, Kentucky, and Goldcoast had an increasing trend during 2006-2019; while the proportion of Salmonella serovars Derby, Senftenberg, Infantis, Newport, Aberdeen, Potsdam, Bovismorbificans, showed a downward trend (Fig. 2).S. Enteritidis was the most common serovar causing human infection in China in 2006 (18.59%) and 2012 (35.90%), while S. Typhimurium became the most common serovar from 2013 to 2018, however, S. Enteritidis became the dominant serovar again in 2019 (Fig. 2).The prevalence of Salmonella serovars Derby, Rissen, Kentucky, Corvallis, Mbandaka, Aberdeen, and Indiana from non-human origin was higher than that of human origin.Moreover, the percentage of S. Typhi isolates that cause typhoid fever has been decreasing in recent years (Fig. S3).Notably, an average of 62 and 46 serovars are detected annually in S. enterica isolates of human and non-human origin, respectively.The number of serovars detected had an increasing trend with fluctuation over time during 2006-2019 (Fig. S4A), and new serovars can be detected annually (Fig. S4B).

Serotyping and multi-locus sequence type based on WGS data
To understand the genomic characteristics and AMR trends of S. enterica isolates, we performed Illumina sequencing on 1962 isolates from 22 Chinese provinces or municipal cities (Figs S5 and S6 and Table S2).The isolate collection represented a wide variety of isolation sources involving five subspecies of S. enterica and a large time span, ranging in date from 1982 to 2019 (Fig. S5A-C).Among the 1962 S. enterica isolates, a total of 130 different serovars were included, S. Typhimurium was the most common serovar, accounting for 25.59% (Table S3 and Fig. S5D).Notably, 29 S. enterica serovars have not been isolated and identified in China before this study, the genomic characteristics of these strains were summarized (Table S4).
Based on WGS data, all 1962 isolates were well serotyped by both traditional serology and WGS using a combination of SeqSero2 and SISTR tool for serovar prediction (100%) (Fig. S5E and Table S5).
In addition to serotyping, multilocus sequence typing (MLST) is an accurate, robust, reliable, high throughput typing method that was also used as an unambiguous tool to examine bacterial relatedness [29][30][31].We found that 200 different STs among the 1962 isolates and 12 previously unknown STs were assigned in seven serovars (16 isolates) (Fig. S8 and Table S6), suggesting that well-known and previously unidentified ST groups are both co-existing and evolving.Generally, ST11 was the dominant ST in the 1962 genomes, followed by ST34 and ST19 (Fig. S9A).Moreover, the diversity of STs has had an increasing trend over time in recent years (Fig. S9B).Among the 1159 isolates of human origin, ST34 was the dominant ST in six Chinese areas, except for North-West (Fig. S9C).Of the 803 isolates of human origin, ST34 was the dominant ST in the South, South-West, Central, and North, while ST11 was the dominant ST in the East (Fig. S9C).
Between 2006 and 2019, the average number of ARGs carried by NTS genomes increased in both human and non-human origins (Fig. 3A  and B), and increased trends were also observed in 4 of 12 antimicrobials of human origin and 7 of 12 of non-human origin (Fig. 3C and D).In particular, the prevalence of beta-lactam resistance genes from 25.93% in 2006-2009 increased to 68.71% in 2014-2017.In general, the proportion of AMR isolates, including beta-lactam, quinolone, tetracycline, and rifampicin which are key antimicrobial agents in human and/or veterinary medicine, have an increased trend over time in both human and non-human origins.The proportion of isolates conferring resistance to phenicol, sulphonamide, and trimethoprim showed a significantly increased trend in non-human origin, while the proportion of sulphonamideand trimethoprim-resistant strains fluctuated and increased in human origin.This phenomenon may be explained by the use of antibiotics (for example, broad-spectrum penicillins, cephalosporins, fluoroquinolones, macrolides, and trimethoprim) in China during the same period, the data of antibiotic use was obtained from the ResistanceMap (https://resistancemap.cddep.org/CountryPage.php?countryId=70&country=China+).
The prevalence of MDR Salmonella isolates across different areas was distinct (Table S13).Chicken and pig are the main sources of nonhuman origin isolates, the proportion of MDR rates also showed an increased trend in different sampling periods (Fig. S10).Positive correlations with resistance to multi-drugs were observed with some serovars.In total, 55.28% (429/776) and 49.29% (452/917) of non-human and human origin NTS isolates had an MDR gene type which was distributed in 39 and 46 serovars, respectively.Among these serovars, Salmonella I 1,4, [5],12: i:-and Choleraesuis exhibited a relatively high proportion of MDR gene types, both exceeding 92.0% (Table S14).However, the proportion of MDR rate in S. Goldcoast of human origin (66.67%) is higher than that of non-human origin (41.67%).
For virulence-associated genes, a total of 129 different virulence factors (VFs) were found (Table S15).No obvious trends or changes in VFs carried by human and non-human origin isolates were observed.Noticeably, the typhoid toxin cdtB was observed in 37 different serovars (Table S16).Most virulence genes are located within highly conserved Salmonella Pathogenic Islands (SPIs) on the chromosome and are usually acquired by lateral gene transfer [32].SPI-1 was the predominant SPI, identified in 99.80% of all S. enterica strains, followed by SPI-2, SPI-3, and SPI-5, accounting for 99.69%, 99.69%, and 99.64%, respectively (Table S17).Salmonella genomic island 1 (SGI1) was identified in 32 isolates (13 serovars) (Table S18).SGI1 is an integrative mobilizable element initially characterized in S. Typhimurium DT104 strains that play an important role in the capture and spread of MDR, subsequently it has been found in clinical isolates of Salmonella serovars and some members of the Enterobacteriaceae [33].The Yersinia high pathogenicity island (HPI) encodes a yersiniabactin-mediated iron acquisition system present in highly pathogenic strains of Yersinia and several members of the Enterobacteriaceae, for example, Escherichia coli, Citrobacter spp., Klebsiella spp., S. enterica subspecies III and VI, and S. Senftenberg [34].To our knowledge, HPI is rare in S. Typhimurium, and this is the first report of HPI in S. Typhimurium strains isolated from humans.
In general, the number of MGEs per isolate was positively correlated with number of ARGs carried in isolates of both human (P < 0.0001, R 2 = 0.4562; Fig. S12A) and non-human origin (P < 0.0001, R 2 = 0.4628; Fig. S12B).However, no correlation was observed between the number of MGEs and VFs per isolate (Fig. S12C and D).
We found that 67.99% (1334/1962) of the isolates contained at least one intact prophage in their genomes (Table S21).A total of 1976 prophages divided into 71 classes were found among the 1962 isolates, with a maximum of four intact prophages observed in a single isolate.Interestingly, Gifsy 2, Salmon 118 970 sal3 and Gifsy 1 were the three most common prophages (n = 537, 490 and 249, respectively).

More ARGs, VFs, and MGEs were observed in S. enterica isolates of non-human origin
We subsequently explored the shared and distinct features of WGS genomes of human and non-human origin (Fig. S14).Overall, the NTS strains isolated from non-human origin possessed more ARGs, virulence genes, MGEs, and AMR resistance phenotype than that from human origin (Figs S15 and S16).Of these WGS isolates of non-human origin, the strains isolated from pigs harbored more ARGs and MDR genotypes than the others (Fig. S17).
We also explored the shared and distinct features of WGS isolates from humans, wet markets, and the environment (Huangpu River water) in Shanghai, which is one of the largest cities in China.Among these serovars, ARGs, MGEs, and plasmids identified in strains of human origin, most of them could be found in strains isolated from wet markets, suggesting that AMR-Salmonella could circulate among human, food (wet markets) and the environment in Shanghai (Fig. S18).Among these WGS isolates, 74.13% (212/287) and 90.30% (763/845) were from blood (Fig. S19A) and diarrhea stool samples (Fig. S19B), respectively.We found that more ARGs and fewer VFs in S. Enteritidis from gastrointestinal infection than that from extra-intestinal infection (Figs S20 and S21).
Moreover, a total of 550 and 8 strains of S. Kentucky and S. Chester were identified in our database, respectively (Figs S22 and S23).Of the 1962 WGS isolates, a total of 10 S. Kentucky strains were included, including five ST198 and ST314, respectively (Table S22).All five S. Kentucky ST198 isolates showed an MDR genotype, with three carrying bla CTX-M-14b and two carrying bla CTX-M-65 , respectively.One S. Kentucky ST198 had an MDR genotype first detected in human fecal samples in Shanxi province.Among the five WGS S. Chester strains, four and one belonged to ST343 and ST411, respectively.Notably, the S. Chester ST343 had been isolated from a foreigner's fecal sample on 16 April 2013 (Table S23).
To further explore the genetic relationship of S. I 1,4, [5],12: i:-ST34 isolates collected from China and other geographical regions, we performed phylogenetic analysis of 157 Chinese isolates and 96 publicly available genomes from different sources, a random subset of isolates that was representative of lineages in recent reports [36,37] (Fig. 6).These results suggested that Chinese isolates were different from those in Southeast Asia, Northwest Europe, Southeast Europe and America, and with a local evolution.Previous studies reported that the global dissemination of the S. I 1,4, [5],12: i:-initially from Europe, then spread to Asia and the USA [36,38].
To investigate the phylogenomic relationship with our dataset of S. Enteritidis ST11 genomes, we generated a maximum-likelihood phylogenic tree inferred from the 6976 single nucleotide polymorphisms (SNPs).From the SNP phylogeny, we found that S. Enteritidis genomes caused human infections mainly associated with chicken, and the majority of S. Enteritidis carrying the bla TEM-1B gene also showed the MDR pattern (Fig. S26).To understand the origins and genetic relationship of S. Choleraesuis, we performed a combined genomic analysis, contextualizing Chinese data with S. Choleraesuis isolates from Europe, Asia and the Americas.The phylogenetic tree shows that the Chinese strains isolated from humans or pigs were closely related to strains isolated from India, Thailand, the United Kingdom and Denmark, but distant from strains from Germany, Vietnam, Brazil and France (Fig. S27).Taken together, these findings suggested that S. Choleraesuis strains collected in this study have various geographic origins and evolved locally.We also explored the relationship between S. Heidelberg genomes recovered from humans and chickens in China and chickens imported from the United States and Brazil.The Chinese strains isolated from chickens were closely related to strains isolated from chickens imported from Brazil, distinct from the strain isolated from a human in China (Fig. S28).It should be noted that the result relies on a single S. Heidelberg isolate of humans for comparison.Typhimurium and its variant I 1,4, [5],12: i:-isolates from humans and food animals in China.Maximum-likelihood tree of 502 S. Typhimurium and monophasic isolates was constructed by using 50 061 SNPs identified by reference to the complete genome sequence of S. Typhimurium strain LT2.The tree is rooted with S. Paratyphi A strain AKU1 12601 complete genome sequence as an outgroup.From the inner to outer circles are   From the phylogenetic tree, we found that 1) clade 1 and 4 only included Chinese strains; 2) clade 2 mainly included Chinese strains and one Australian strain which were closely related to 2 and 3 strains from Canada and Denmark, respectively; 3) clade 3 were Chinese isolates and were clustered with those isolated from pigs in Japan; 4) clades 5 and 6 included Chinese strains and strains isolated from cattle from Japan.In addition, two Chinese isolates from humans clustered with one American isolate from pig and Australian strains from humans.
To understand the population structure currently causing typhoid in China, we constructed a maximum-likelihood phylogeny of 64 strains of S. Typhi (Fig. S29).The results based on 8772 SNPs showed the division of isolates into two clades, and both of them were observed in Xinjiang.A phylogenetic analysis of Paratyphi B and its variants revealed that I 1, 4, [5], 12: b:-and S. Paratyphi B var. Java ST42 genomes recovered from humans clustered together (Fig. S30).Moreover, the isolates Paratyphi B ST86 and Paratyphi B var. Java ST42, ST43, ST423 from aquatic animals were clustered with humans.These results indicated that aquatic animals are potential host and transmission vectors of Paratyphi B ST86 and its variants.

DISCUSSION
To our knowledge, this is the largest collection and first systematical study reporting the temporal and spatial dynamics of S. enterica serovars across 23 provinces or municipal cities over 14 years in China.S. Typhimurium (26.82%) is the most common serovar causing human infection in China (Fig. 1B), which was similar to Africa [6], Belgium [39], and Australia [7], and inconsistent with other high-income countries and regions [9,40].Salmonella spp. was the second primary (6.8%) bacterial pathogen of foodborne disease outbreaks in China, many of which were linked to the consumption of animal source foods, which were considered to be the major reservoirs from which Salmonella was spread [41,42].Moreover, our data also showed that Salmonella serovars Enteritidis, Derby, Typhimurium, Thompson, and Aberdeen were the most common serovars detected in chicken, pig, duck, aquatic product, and turtle, respectively, which is consistent with the findings in previous reports [42,43].
Among the 502 WGS S. Typhimurium strains registered at the Bacterium-learning Union, 32.67% of strains (n = 164) were predicted as S. I 1,4, [5],12: i:-, indicating WGS can help us better distinguish S. 1,4, [5] : 12 : i : 1,2 and S. I 1,4, [5],12: i:-.Previous studies have shown that incongruent matches were present between conventional serotyping and WGS identification using SeqSero2 or SISTR in S. enterica strains (for example, serovar variants) from foods and human beings [27,[44][45][46].However, these problems could partly be overcome by using WGS in combination with in silico analysis tools.Notably, a total of 29 serovars that have not previously been reported in China were identified by conventional serotyping and WGS prediction in this study.Briefly, this is the largest collection of conventional serotyping results by using WGS prediction Salmonella serovars (a total of 130 serovars) in China.A total of 164 serovars were identified in China in this study, however with the rapid development of the economy and global trade, and the aging of human beings, the risk of foodborne pathogens' dissemination will further increase, and more and more Salmonella serovars will be detected in China.
Phylogenetic analysis revealed that ST34 from pig and ST19 from chicken origin were mainly associated with isolates causing child and adult gastroinfections, respectively.The previous study reported by the CDC and EFSA also emphasized that pig and pork meat are the main sources of S. Typhimurium and its variant [16].Significant differences were observed in Chinese S. I 1,4, [5],12: i:-ST34 strains than strains in Southeast Asia, Northwest Europe, Southeast Europe and America.These results showed that a local evolution was present in Chinese strains.
For most drugs, the detection of resistance genes was both sensitive (>98%) and specific (>99%) in predicting AMR phenotype [24][25][26][27].WGS revealed a diversity of known genetic mechanisms for AMR that accounted for >45% of isolates that had an MDR genotype.From 2006 to 2019, the proportion of AMR S. enterica strains showed an increasing trend in both human and non-human origin (Fig. 3C and D).Moreover, we also found the AMR rate of most antimicrobials of human origin isolates decreased in 2012, which was probably influenced by the policy for antimicrobial management (http://www.gov.cn/flfg/2012-05/08/content2132174.htm).A recent study reported that the prevalence of resistance in Salmonella showed an increased trend after 2010 in penicillin in chicken, and in sulfonamide, penicillins, and tetracycline in pigs [22].A similar landscape was also observed in other countries, for example, the prevalence of MDR was generally low ranging from 3.4% in 1984 to a high of 9.7% in 2013 in Australia [7].Although some changes in antimicrobial usage policy in China during this timeframe, including human medicine [47] and animal husbandry (http://www.gov.cn/gongbao/content/2013/content 2528114.htm,http://www.moa.gov.cn/govpublic/ncpzlaq/201507/t20150723 4759838.htm), it is also easy to obtain antimicrobials without a prescription in retail pharmacies [48].In some serovars, the prevalence of MDR was greater than 50%, while the serovars were different from that in Australia [7].In addition, more AMR genes and MGEs were identified in strains such as S. 1,4, [5] : 12 : i : 1,2 and S. I 1,4, [5],12: i:-isolated from non-human rather than human origin, suggested a possible transmission route from food animal to humans through the food chain.
Our findings highlight the presence of bla CTX-M-55 , bla CTX-M-14 , bla CTX-M-65 , bla CTX-M-123 , bla CTX-M-14b , bla CTX-M-15 , and bla CTX-M-130 , which were the most common CTX-M type genes detected in patients with diarrhea [42].This is the first study reporting the detection of the bla CTX-M-130 gene in S. Paratyphi B recovered from a 10-month-old boy's fecal sample.Overall, these results improved our understanding of the transmission dynamics of CTX-M genes, which are known to be carried on plasmids, between animals, humans and environments.Of the 775 beta-lactamase-producing isolates, 30.45% (n = 236) contained at least one QRDR.Our findings suggest that the incidence of PMQRs in S. enterica isolates is increasing (Fig. 3C  and D), as previously reported [42].Moreover, 12 QRDR point mutations, 6 in GyrA, 2 in GyrB, 3 in ParC, and 1 in ParE were found in 1537 isolates.
However, the emergence and rapid spread of mobile resistance genes mechanisms MCR-1 [49] and tet(X4) [50,51] which have gained global attention.In this study, a total of 27 S. enterica isolates were positive for mcr genes, accounting for 1.38% of all WGS isolates.However, the prevalence of mcr-1 was 0.76% and 1.19% in 2010-2013 and 2014-2017, respectively.After the cessation of colistin use as a feed additive for animals in China from 30 April 2017, the prevalence of mcr in Salmonella may show a decreasing trend in the future.Similar to previous reports [52], S. Typhimurium and its variant was the most common serovar carrying mcr genes, accounting for 29.63% (8/27) and 33.33% (9/27), respectively.A recent genomic epidemiological study reported that serovars and STs were associated with a high rate of colistin resistance among S. enterica isolates [53].Among these mcr-positive isolates, 10 STs including 1 novel ST were identified, with ST34 the predominant ST.We reported the identification of mobile tigecycline resistance gene tet(X4) in an S. enterica serovar Llandoff ST8300 isolate, which is a potential host for the spread of tigecycline resistance.
Recent studies [54,55] reported that a relatively high prevalence of fosA7 (>9.00%) was identified in Salmonella isolates from food animals and retail meat products in China, which was consistent with our results.We also found that 2.9% of Salmonella isolates of human origin carried fosA7.To our knowledge, this is the first study that comprehensively describes the prevalence of fosA7 in Salmonella from human and non-human origin sources in China.
Inappropriate use of antimicrobials in livestock production and human clinical medicine may contribute to the development and accumulation of AMR in pathogenic bacteria and is a major threat to public health.A recent study that compared AMR with antimicrobial usage data showed that the preva-lence of resistance to fluoroquinolones and thirdgeneration cephalosporins was correlated with usage [56].Moreover, the broad patterns of antimicrobial use in the treatment of diarrhea across the Global Enteric Multicenter Study sites also showed that macrolides (azithromycin) and fluoroquinolones (in particular ciprofloxacin) were used more frequently in Asia [57].
Our work has several limitations.First, the dataset used in this study consisted of data collected for different reasons, during a varying time period and with different projects in mind.We cannot rule out sampling bias because the comprehensive network system is not built up yet, however, as far as we know, it is currently the largest database for Salmonella in China.We acknowledge the potential bias, which needs to be studied further.Among the strains isolated from patients and farm animals, we did not obtain a history of drug use.The previous study indicated higher rates of antibiotic resistance in high consuming countries [58].Second, no data were available on total meat consumption including all meat types each year by provinces or municipal cities.Third, as for the 29 serovars mentioned above which have not been reported in China and their correlation with strains isolated from other countries were not explored.Fourth, no phenotypic data of antimicrobial susceptibility testing were available, AMR trends were explored based on WGS prediction.Another limitation is that we did not sequence all the 164 serovars of S. enterica strains in the Bacterium-learning Union, but selected 130 serovars of them.Moreover, S. 1,4, [5] : 12 : i : 1,2 and S. I 1,4, [5],12: i:-were not well distinguished in our database, we will refine this shortcoming in the next study.Last, we collected 35 382 S. enterica isolates from 23 provinces or municipal cities, but only provide WGS genomic data for 5.55% of the isolates.Moreover, WGS could not determine the precise location of ARGs, VFs, and MGEs, so long-read sequencing methods are needed in the future.

CONCLUSION
In conclusion, our work provides the first systematic overview of current serovar prevalence and AMR trends of S. enterica strains isolated from human and non-human origins in China.S. Typhimurium, S. Enteritidis, and S. Derby were the main serovars responsible for human infection.Moreover, we have presented a most detailed genomic study of S. enterica in China during 2006-2019.Up till now, there is no comprehensive national surveillance scheme involving humans, food, animals, and the environment in China for continually monitoring AMR and geographical distribution of S. enterica.To our knowledge, our findings represent the largest longitudinal surveillance system for S. enterica in China and provides valuable public health knowledge on the trends and distribution of serovars and AMR in S. enterica.This large, observational study will provide valuable information for food safety interventions, control, and prevention of the development and spread of AMR Salmonella enterica in China.

Dataset and study design
In this study, a total of 35 382 S. enterica isolates that had been obtained from various sources and geographical areas were used and analyzed.The detailed description of dataset and study design is summarized in Supplementary materials and methods.In order to explore the temporal and spatial dynamics of serovars and AMR, the correlation of strains between humans and food animals, the distinct features of serovars of gastro-and extra-intestinal infection of S. enterica, and evaluate the accuracy of traditional serology, we conducted a WGS study.Our final WGS data set comprised 1962 S. enterica isolates including human (n = 1159) and non-human origin (n = 803) from 22 provinces or municipal cities, ranging in date from 1982 to 2019 (Fig. 1, Fig. S5 and Table S2).

WGS and gene content analysis
All 1972 S. enterica strains were subjected to WGS after cultivation for 18 h at 37 • C and 180 rpm.Raw reads were processed with Trimmomatic version 0.36 [59], Unicycler version 0.4.7 [60], and Prokka version 1.13.3 [61], respectively, to generate highquality reads, assembled contigs, and annotation of genome sequences.Individual accession numbers for assembled Illumina sequence data are available in Table S24.

Statistical analyses and visualization
Statistical significance was taken at P < 0.05.Multiple comparisons were performed by the Kruskal-Wallis test and Mann-Whitney U test (unpaired t-test) using GraphPad Prism version 8.0.Venn diagrams were drawn in Venn Diagrams (http://bioinformatics.psb.ugent.be/webtools/Venn/).Easyfig (http://easyfig.sourceforge.net/)[79] was used to visualize the genetic context comparisons.The detailed description of materials and methods is summarized in Supplementary materials and methods.

DATA AVAILABILITY
All assembled Illumina sequence data are available from the National Center for Biotechnology Information under the BioProject number: PRJNA766315.Sequence reads generated for this study have been submitted to the National Microbiology Data Center (NMDC) under BioProject number: NMDC10017893 and NMDC10018145.All data and database in this study are publicly available, the detailed description of data availability is summarized in Supplementary materials and methods.

Figure 2 .
Figure 2. Temporal trends of S. enterica serovars during 2006-2019 in China (top 30 serovars).TH: The total number of Salmonella isolates of human origin.TNH: The total number of Salmonella isolates of non-human origin.TT: The total number of Salmonella isolates of non-human origin.

Figure 3 .
Figure 3. Temporal and spatial changes in AMR of NTS in China.A and B, The distribution of numbers of ARGs in each NTS isolate from both human origin (A) and non-human origin (B) grouped by sampling time.(Both, n = 1713; Human origin, n = 937; Non-human origin, n = 776).The mean ARG count for each period is noted.Multiple comparisons were performed by the Kruskal-Wallis test using GraphPad Prism version 8.0.The ' * ' on the right represent P-values.* : P < 0.05; * * : P < 0.01; * * * : P < 0.001.ns, no significant difference.C, Temporal dynamic of AMR genotypes in NTS isolates from human origin (n = 937).D, Temporal dynamic of AMR genotypes in NTS isolates from non-human origin (n = 776).

Figure 4 .
Figure 4. Temporal changes in the prevalence of ARGs among NTS isolates.Only significant temporal trends and critically important resistance genes are shown (1% significance level).Colored bars represent the predicted phenotype conferred by ARGs, and they are clustered by their resistance to different classes of antibiotics.ARGs that were detected at >1% prevalence in NTS genomes are shown (n = 1710).

Figure 5 .
Figure 5. Phylogenetic analysis of S. Typhimurium and its variant I 1,4,[5],12: i:-isolates from humans and food animals in China.Maximum-likelihood tree of 502 S. Typhimurium and monophasic isolates was constructed by using 50 061 SNPs identified by reference to the complete genome sequence of S. Typhimurium strain LT2.The tree is rooted with S. Paratyphi A strain AKU1 12601 complete genome sequence as an outgroup.From the inner to outer circles are 1. Isolation source, 2. Province, 3. Year, 4. Serovar, 5. Age group, 6. ST of the isolates.