Horizontal Plasmid Transfer Promotes the Dissemination of Asian Acute Hepatopancreatic Necrosis Disease and Provides a Novel Mechanism for Genetic Exchange and Environmental Adaptation

Global outbreaks of shrimp acute hepatopancreatic necrosis disease (AHPND) caused by V. parahaemolyticus represent an urgent issue for the shrimp industry. This study revealed that the transmission mode of AHPND consists of two steps, the transregional dissemination of V. parahaemolyticus and the horizontal transfer of an AHPND-associated plasmid. Surprisingly, the introduction of the AHPND-associated plasmid also offers a novel mechanism of genetic exchange mediated by insertion sequences, and it improved the fitness of V. parahaemolyticus in a harsh environment. The results presented herein suggest that current shrimp farming practices promote genetic mixture between endemic and oceanic V. parahaemolyticus populations, which introduced the plasmid and accelerated bacterial adaptation by the acquisition of ecologically important functions. This entails a risk of the emergence of new virulent populations both for shrimp and humans. This study improves our understanding of the global dissemination of the AHPND-associated plasmid and highlights the urgent need to improve biosecurity for shrimp farming.

dustry. This study revealed that the transmission mode of AHPND consists of two steps, the transregional dissemination of V. parahaemolyticus and the horizontal transfer of an AHPND-associated plasmid. Surprisingly, the introduction of the AHPND-associated plasmid also offers a novel mechanism of genetic exchange mediated by insertion sequences, and it improved the fitness of V. parahaemolyticus in a harsh environment. The results presented herein suggest that current shrimp farming practices promote genetic mixture between endemic and oceanic V. parahaemolyticus populations, which introduced the plasmid and accelerated bacterial adaptation by the acquisition of ecologically important functions. This entails a risk of the emergence of new virulent populations both for shrimp and humans. This study improves our understanding of the global dissemination of the AHPND-associated plasmid and highlights the urgent need to improve biosecurity for shrimp farming.
KEYWORDS Vibrio parahaemolyticus, environmental adaptation, genetic exchange, insertion sequence, transmission mode V ibrio parahaemolyticus is a Gram-negative, halophilic bacterium that is widespread in warm estuarine and marine environments (1). As an important foodborne pathogen, V. parahaemolyticus is becoming the leading cause of acute gastroenteritis due to the increased consumption of raw or undercooked seafood. Recently, V. parahaemolyticus has gained particular notoriety because it causes massive acute hepatopancreatic necrosis disease (AHPND) in shrimp (1,2). This disease has resulted in economic losses of over $50 billion in global shrimp aquaculture (3). A recent study suggested that V. parahaemolyticus can cause AHPND symptoms in shrimp due to the acquisition of a 70-kb plasmid encoding the binary toxin PirAB vp (4). Evidence of the transfer of pirAB-bearing plasmids between different Vibrio species has been found (5,6). The pirAB genes and their flanking genes form a mobile genetic element (MGE) called the pirAB-Tn903 composite transposon, or Tn6264 (6); it consists of six genes and two identical insertion sequences named ISVal1.
AHPND was first identified in China and Vietnam in 2010 (7,8), and it was reported in Malaysia and Thailand in 2011 (9,10). Afterwards, it was subsequently detected in Mexico and South America (11,12). The transmission of V. parahaemolyticus was speculated according to this timeline, which has caused long-term disputes in the international shrimp trade between Asia and Mexico. However, our previous genomic study revealed that multiple lineages of V. parahaemolyticus have emerged independently worldwide, with no clear patterns of transmission (2), which is inconsistent with the massive cyclic dysentery epidemics reported in Asia and other countries around the world. To date, very little is known about the origins and spread of V. parahaemolyticus in shrimp farming regions. As farmers often sell diseased shrimp with a high load of the pathogen in the market to reduce economic loss, massive AHPND outbreaks may also pose a threat to public health.
A recent study revealed the existence of binary toxin PirAB vp -bearing Vibrio spp. isolated from shrimp long before 2010 (12), suggesting the possibility of long-term concealment of AHPND-causing V. parahaemolyticus (Vp AHPND ) strains, without an epidemic outbreak. In our recent study, we found Vp AHPND in shrimp most likely introduced from the sediment (13). However, it is still unclear how environmental V. parahaemolyticus strains emerged and adapted to the environment, resulting in shrimp and human disease.
The aim of this study is 2-fold. First, we aimed to understand why multiple lineages of V. parahaemolyticus rather than a single clone emerged in AHPND outbreaks. To this end, a set of V. parahaemolyticus isolates from humans, shrimp, and the environment were selected from the shrimp farming region. Whole-genome analysis was then carried out to reconstruct the spatial and temporal spread of V. parahaemolyticus and to understand whether horizontal plasmid transfer occurred among different V. parahaemolyticus populations. Second, we aimed to determine whether frequent genetic exchanges among V. parahaemolyticus populations were mediated by an insertion The SNP mutation rate estimated by BEAST yielded a substitution rate of 1.40 ϫ 10 Ϫ6 site Ϫ1 year Ϫ1 (95% highest posterior density [HPD], 1.29 ϫ 10 Ϫ6 to 1.51 ϫ 10 Ϫ6 ) or 2.9 SNPs per genome per year for the ST415 clone. The relaxed clock models and strict clock models yielded nearly identical clock rates. Hence, our results are in line with the strict clock model with a constant population size. The molecular rate is similar to that observed in the V. parahaemolyticus ST36 clone (15). Divergence time analysis of 19 ST415 strains classified them into three stages. BEAST analysis indicated that the most recent common ancestor of ST415 was in February 1996 (95% HPD, May 1995 to October 1997), which is consistent with the time of the first identification of ST415 in Thailand according to the PubMLST database (in 1996). These results suggest that ST415 clones had been established over a long period of time before the AHPND outbreak. By May 1997, ST415 was transmitted to Zhangpu, Fujian Province (lineage II). Afterwards, ST415 spread into East China and diversified (lineage III), as there were a few hundred SNPs on the branch. This stage was estimated to have occurred around October 2000. ST415 subsequently spread to North China by April 2011. Therefore, this ST likely spread as a pandemic clone, resembling the dissemination of the ST3 pandemic in early 1996.
We further extracted and sequenced the plasmids from 19 ST415 genomes. pirABpositive plasmids were identified in all ST415 genomes except for ZP8. Interestingly, sequence analysis showed that the pirAB-positive plasmids from 10 ST415 strains, as well as plasmids pVpR14, pHZ52-1, and pVPWZ2, harbored one or two redundant ISVal1 sequences. Together with the above-mentioned plasmids, the pirAB-positive plasmids can be divided into four clusters (Fig. 2B). Overall, the clustering of the plasmids largely reflects the evolutionary relationships of the ST415 chromosomes, except for the plasmids from the Vietnam and Jiangmen (China) strains, which clustered with those obtained from northern China. Genetic analysis showed that redundant ISVal1 sequences are inserted in different positions in Tn6264 and that they can be divided into six different variants (Fig. 2B). In addition to two flanking ISVal1 sequences, Tn6264 consists of six genes (namely, genes 1 to 6). Variant 1 harbors an additional ISVal1 (gene 7) and a hypothetical protein (gene 8) inserted between genes 1 and 2 (pirB) and an additional ISVal1 inserted between genes 5 and 6. In contrast to variant 1, variant 2 only has one ISVal1 inserted between genes 5 and 6. Variant 3 has a second redundant ISVal1 inserted between genes 7 and 8. For variant 4, gene 1 was replaced by ISVal1 and by another hypothetical protein (gene 9). Interestingly, pVpR13 (variant 5) also harbors a redundant ISVal1 inserted next to pirA (gene 3). It seems that ISVal1 has been randomly inserted into different positions in Tn6264 and that it shaped the genome plasticity of pirAB-positive plasmids. In the next section, we further examine the transferability of ISVal1 from plasmid to chromosome. The variability of Tn6264 in ST415 strains also suggests that pirAB-positive plasmids become genetically diverse during the transmission of ST415 strains.
Likewise, the genetic analysis of AHPND-associated plasmid in Vp AHPND strains from Southeast (SE) Asia also suggested that the emergence of ST970 Vp AHPND strains facilitated the horizontal transfer of plasmids among different STs in SE Asia, thus promoting the dissemination of AHPND endemically (Text S1 and Fig. S2). Phylogenetic analysis of 88 assembled pirAB-positive plasmids revealed extensive plasmid transfer among the common STs found in this study (Fig. S3). This enabled us to reconstruct the historical dissemination of AHPND-associated plasmids in Asia assisted by the transmission of ST970 (transmission route 1 [T1]), ST415 (T2 and T4 to T7), and ST1166 (T3) (Fig. S4).
Horizontal plasmid transfer promoted the genetic exchanges of V. parahaemolyticus. Next, we conducted 3 years of consecutive monitoring in a shrimp farming region to observe the plasmid transfer and its subsequent consequences in vivo. In December 2014, strains of ST452, ST1803, and ST978 were isolated from the sediment of shrimp ponds that had not been used for rearing shrimp (Data Set S3). The rearing of postlarval shrimp started in May 2015. No V. parahaemolyticus was detected in the postlarval shrimp before the introduction of shrimp into the pond. Thereafter, we revisited the farm after the suspected AHPND outbreaks occurred (June 2015). Strains HZ14 (ST415), HZ15 (ST1803), and HZ18 (ST978) were identified in the shrimp. These observations suggest that ST1803, ST978, and ST452 are likely to be the endemic STs, while ST415 was introduced to the shrimp farm after 2014. However, during the sampling in June 2016, suspected AHPND outbreaks occurred again. The genotypes in the shrimp pond became genetically diverse. A total of 12 strains were isolated in 2016. MLST subtyped them into ST415, ST452, ST978, ST1803, and an undefined ST with no PCR amplification for the gene dtdS (encoding threonine dehydrogenase). ST415 and undefined ST strains both harbored pirAB-positive plasmids, which carried a redundant ISVal1. Phylogenetic analysis of their chromosomes found that two ST452 strains clustered with two undefined ST strains (Fig. 3A), indicating that this undefined ST might have originated from ST452.
We also analyzed the plasmid profiles for the sequenced strains in different years. In 2014 and 2015, the strains from ST452, ST1803, and ST978 all harbored a 106-kb pirAB-negative plasmid (plasmid 1, Data Set S3). However, since 2015, when the HZ14 strain (harboring a 69-kb pirAB-positive plasmid [plasmid 2] and a 76-kb pirAB-negative plasmid [plasmid 3]) was identified, the plasmid profiles of ST452 have changed. Afterward, plasmids 2 and 3 were transferred from ST415 to ST452 with a dtdS deletion, while normal ST452 only acquired plasmid 3. The ST1803 and ST978 strains still only harbored plasmid 1 (Data Set S3).
Remarkably, the genomic analysis of two undefined ST strains showed that the insertion of ISVal1 resulted in the loss of an 11.3-kb DNA fragment enclosing the dtdS locus (Fig. 3B). In addition, the insertion of ISVal1 was identified in various locations of the chromosomes of strains HZ52 and HZ49 as well as in their plasmids. Further analysis showed that a total of 24 ISVal1 insertions were found throughout the genome of strain HZ52, including a replacement event, five deletion events, and two insertion events (Table 1).
In strain HZ52, a 54-kb segment encoding the YefM-YoeB toxin-antitoxin system, acetyltransferase, and proteins encoded a few genes associated with metabolism was replaced by a 90-kb genomic island (Fig. 3C). This genomic island was flanked by the ISSod13 transposase and integron integrase IntI4 with three ISVal1 copies inserted. Interestingly, this genomic island can be divided into three sections. The first section was found in the chromosome of strain HZ-7 (isolated from another shrimp farm in Hangzhou), while the second section separated by ISVal1 was located in a plasmid from HZ-7. However, the origin of the third section remains unclear. Thus, it is reasonable to speculate that ISVal1 might mediate the excision of the first section, be incorporated with the second section, and transferred as a whole into HZ52. However, a gene-bygene analysis showed that this genomic island is functionally similar to the replaced segment, as the two harbored a similar set of genes. The inserted genomic island carried an additional toxin-antitoxin system.
The five deletion events in strain HZ52 resulted in the loss of genes encoding a TonB-dependent receptor, dtdS and its surrounding region, acetyltransferase/longchain fatty acid transport protein/ADP-ribose pyrophosphatase, endonuclease I, and Flp pilus assembly. An insertion event introduced genes encoding a type I restrictionmodification system. In addition, plasmid 3 in strain HZ49 harbored an additional 12-kb drug-resistant island (DRI) flanked by two ISVal1 copies. Five antibiotic resistance genes were identified in the DRI, aadA16, ARR-3, sul1, tet(B), and dfrA27. This DRI has also been found in the plasmid pVPSD2016-2 from strain 237 but without ISVal1, indicating that ISVal1 might facilitate the formation of composite transposons and be transferred between plasmids.
Likewise, we also found a 127-kb genomic island in strain 201 and another ST1166 Vp AHPND strain, which was also possibly associated with the introduction of the pirABpositive plasmid (Fig. 3D). This genomic island harbored three ISVal1 copies that formed another composite transposon.
To confirm whether ISVal1 can move between plasmids and chromosomes, an in vitro evolutionary experiment was conducted. A plasmid containing two ISVal1 copies was constructed and cocultured with V. parahaemolyticus strain ATCC 17802 for 10 days. At the end of the experiment, the disappearance of the plasmid occurred when sacB (located on the plasmid) was not detected through PCR (Table 2). At day 10, a total of 58 colonies were randomly selected for PCR detection of ISVal1 and sacB. Four out of 58 colonies were ISVal1 positive and sacB negative, indicating that the insertion rate of ISVal1 is 6.9%. One of the colonies was selected for genome sequencing. The results confirmed that ISVal1 was transferred from the constructed plasmid ( BLASTn results show that insertion of ISVal1 occurred in the chromosomes of 50 Vp AHPND and 5 AHPND-associated Vibrio sp. strains, with the number of ISVal1 ranging from 1 to 25 (Data Set S4), but this did not occur in any non-Vp AHPND strains. These observations indicate that the insertion of ISVa11 into the chromosome was caused by the introduction of a pirAB-positive plasmid, which provided an underlying mechanism for the transfer of MGEs and virulence genes.
Phenotypic difference between strains HZ52 and HZ79 reveals the role of genetic exchange in environmental adaptation. Next, we tested the phenotypic  difference between strains HZ52 and HZ79 to determine whether the insertion of ISVal1 contributed to environmental adaptation. The growth assay showed that there was no significant difference between strains HZ52 and HZ79 in terms of growth rate (Fig. 4A). Likewise, strain ATCC 17802 with the insertion of ISVal1 also had a growth rate similar to that of its wild type. The growth rates between HZ52 and ATCC 17802 were also not significantly different. In addition, strains HZ52 and HZ79 exhibited similar swimming ability (P Ͼ 0.05, Fig. 4B), indicating that the loss of the scr and csu operons was not determinative for mobility. Biofilm formation assays showed that the biomass of biofilms for strain HZ52 was significantly greater than that for strain HZ79 formed on chitin flakes and glucose on days 2 and 3 (P Ͻ 0.01, Fig. 4C). In addition, strain HZ52 exhibited 4.2-fold higher cpsA expression relative to HZ79 (P Ͻ 0.001), indicating higher capsular polysaccharide (CPS) production (Fig. 4D). To determine whether HZ52 also exhibited a higher biofilm formation ability under continuous-flow and -grazing conditions, V. parahaemolyticus strains and ciliate Coleps hirtus were coincubated in flow cells with chitin. The grazing effects were determined with the quantification of V. parahaemolyticus CFU and the numbers of C. hirtus in the biofilm effluent. The results suggest that with or without grazing pressure, effluents from strain HZ52 both contained approximately 1.5-fold more V. parahaemolyticus cells than did the effluents from strain HZ79 (P Ͻ 0.01, Fig. 4E). Direct microscopic enumeration also revealed that the number of C. hirtus cells increased in the flow cell system inoculated with strain HZ79, relative to the one inoculated with strain HZ52 (Fig. 4F) on day 3, although the differences between the two groups were not significant.

DISCUSSION
Horizontal plasmid transfer promoted the genetic exchange and environmental adaptation for Vp AHPND strains. In this study, we provide the first global analysis of the transmission pattern of Vp AHPND strains by genomic epidemiology. The results herein suggest that the transmission of a few genotypes of Vp AHPND strains is an important driving force promoting the transfer of pirAB-positive plasmids and AHPND. The temporal and spatial analysis of ST415 strains suggests that ST415 is possibly another pandemic clone that was transmitted simultaneously with the pandemic ST3, which has been transmitted along the coastline of China in sequential order from 1996 to 2010 (16). However, due to insufficient sequenced strains from the Americas, it is still not clear how AHPND spread from Asia to South America. Another limitation of this study is that horizontal plasmid transfer events were only epidemiologically observed at one site. Nevertheless, given the observations by Dong et al. (17), plasmid transfer among different Vibrio spp. is likely. Extensive plasmid transfer events seemed to occur during the dissemination of ST415, ST1166, and ST970, which greatly promoted the spread of AHPND in Asia and accelerated genetic mixing. With the exception of the above-mentioned three STs, the STs remained endemically present in each region and became Vp AHPND strains after receiving the plasmid.
Unexpectedly, the introduction of the pirAB-positive plasmid also promoted genetic mixing within V. parahaemolyticus populations caused by several insertion and deletion events. Observation on a shrimp farm showed that pirAB-negative (avirulent) strains were present in the environment until the introduction of ST415. Afterwards, ST452 strains acquired both pirAB-positive and pirAB-negative plasmids from ST415 strains present in the local environment. Thereafter, ISVal1 excised from the pirAB-positive plasmid, resulting in three types of genetic changes for ST452 strains (Fig. 5), including an 11-kb deletion in the region surrounding dtdS. These observations suggest that the transmission of V. parahaemolyticus and subsequent horizontal plasmid transfer might confer upon ISVal1 the ability to mediate the transfer of MGEs. It seems that ISVal1 excised from the plasmid to the chromosome by replicative transposition instead of cut-and-paste transposition as in the pirABpositive plasmid from ST415 and ST452 harbored the redundant ISVal1. Along with plasmid excision, transposons are faced with the task of locating new insertion sites in the genome to integrate within. Some transposon elements are very strict in the sequence feature, such as the Tc1/mariner elements, which always integrate into a TA dinucleotide (18). However, the insertion sites for ISVal1 are more flexible and vary in nucleotide composition, and no integration site preference was observed (Data Set S4). This feature enables ISVal1 to be a powerful gene vehicle and to insert into both the plasmid and the chromosome.
Among these insertion events, the most interesting one is the deletion of the 11-kb region including the dtdS gene, scr operon, and csu operon which is mostly associated with cell swimming ability. Therefore, HZ52 might have a defective swimming function. However, Boles and McCarter suggested that mutations in the scrABC genes greatly reduce but do not eliminate lateral flagellin production; conversely, the loss of the scr operon enhanced the production of CPS (19). Therefore, these genes are not absolutely required for cell mobility. In addition, because lateral flagellar production is a costly process that is not beneficial for the cells that are already on the surface, the deletion of the 11-kb segment might control lateral gene expression and enhance CPS production. To test this idea, the expression of a CPS synthase gene (cpsA) was monitored in HZ52 and HZ79, and HZ52 showed significant higher cpsA expression (P Ͻ 0.01). Further validation experiments supported the above-mentioned hypothesis. Compared with HZ79, strain HZ52 showed slightly lower cell mobility but significant higher biofilm formation ability under both static and continuous-flow conditions. Thus, strain HZ52 might have better survival in the shrimp pond. Therefore, genetic exchanges mediated IS IS ISVal1 Val1 IS

IS
Step 1 Horizontal plasmid transfer Step 2 Replicative transposition Step 3 Three types of genetic exchanges by ISVal1 promoted the insertion and deletion of the genomic island, which might enhance the environmental adaptation of endemic V. parahaemolyticus populations and turn avirulent strains into virulent strains that cause disease in both shrimp and humans. Transmission mode of AHPND and its implications for shrimp and human disease management. Our results also suggest that the transmission mode accounting for the occurrence of AHPND has great impacts on shrimp and human disease management (Fig. 6A). In this mode, the endemic V. parahaemolyticus population is pirAB free. However, as the majority of shrimp ponds are located in estuary regions with the frequent exchange of oceanic Vibrio spp., pirAB-positive Vibrio spp. might have recently entered the shrimp ponds and thus be responsible for AHPND outbreaks or may have been carried by postlarval shrimp to other regions (20).
This mode provides significant implications for shrimp and human disease management. Epidemiological investigation showed that a large quantity of mangroves were destroyed to make shrimp ponds prior to 2010 in SE Asia and South China, which greatly increased the exposure of shrimp to opportunistic pathogens from the ocean (Fig. 6B). Such farming practices in the coastal region not only disrupted the ecological systems but also increased genetic mixing between pathogenic and nonpathogenic Vibrio spp. This aquaculture mode created an environment in which the opportunities of contact between the populations increased, thus maximizing the probability of genetic material transfer and recombination, as observed in Vibrio vulnificus (21). Genetic exchange (via plasmids and genomic islands) at this shrimp farm also presents a good example of how the transfer of antibiotic resistance has taken place between plasmids. He et al. reported that multidrug-resistant V. parahaemolyticus strains were frequently identified in shrimp from Jiangsu Province, China, and suggested that MGE is the main vehicle for the transfer of resistance (22). This study also highlights that the formation of composite transposons mediated by ISVal1 provides a novel mechanism for antibiotic resistance transfer.
From a One Health perspective, current shrimp farming practices resulting in widespread Vp AHPND strains also pose an emergent threat to public health. Current shrimp farming practices have promoted frequent exchange with oceanic vibrios, which could entail a risk of the emergence of virulent populations, with potentially devastating consequences for both aquaculture and human health. By incorporating shrimp, environmental strains, and clinical strains, three farm-to-table spread events were identified. This study also confirmed the transmission of Vp AHPND strains from the environment to the farm and to the table genomically. A more comprehensive spread network should be investigated for other SCGs. To minimize loss, diseased shrimp with a high load of vibrios are often shipped to market, resulting in a possible increase in human infections. Thus, improving biosecurity management may be the key to minimizing the progression of AHPND and reducing subsequent human infection risks. The transformation from traditional pond farming to indoor recirculating aquaculture systems would be a promising direction for effective disease control and prevention, as this new aquaculture mode promotes rigorous biosecurity management, which could efficiently prevent the entrance of external pathogens.
The results presented herein suggest that the transmission of ST415 Vp AHPND strains, as well as of a few other STs, promoted horizontal plasmid transfer and turned endemic V. parahaemolyticus into virulent Vp AHPND . This study identified a novel genetic exchange mechanism mediated by ISVal1. Current shrimp farming practices have promoted frequent exchange with oceanic vibrios, which could entail a risk of the emergence of virulent populations with potentially devastating consequences for both aquaculture and human health. This study addressed the basic question regarding the origins and evolutionary history of Vp AHPND , with significant implications for shrimp and human disease management, and highlighted the urgent need to improve the biosecurity of shrimp from the farm to table. (a) Mangrove was destroyed to give a way for shrimp farming.
(b) High frequency genetic exchanges between endemic and oceanic vibrios Transfer of plasmid, genomic island and antibiotic resistance.
(c) Diseased shrimp with a high load of vibrios is shipped to market to minimize the loss.

MATERIALS AND METHODS
Bacterial isolates and total DNA extraction. The V. parahaemolyticus isolates analyzed in this study are listed in Data Set S1. Total DNA was extracted from the overnight culture using the Wizard genomic DNA kit (Promega).

Sample collection and bacterial isolation and identification from a shrimp farm in Hangzhou.
In December 2014, May 2015, June 2015, and June 2016, samples of sediment, postlarval shrimp, and diseased shrimp were collected from a shrimp farm in Hangzhou, China. Hepatopancreases from shrimp and sediments were aseptically disaggregated in 100 ml alkaline peptone water (peptone, 10 g per liter; sodium chloride, 10 g per liter) and streaked on thiosulfate-citrate-bile salts-sucrose (TCBS) plates, which were incubated at 28°C for 12 h.
MLST. In silico MLST typing of a publicly available V. parahaemolyticus genome was performed using the MLST 2.0 server from the Center for Genomic Epidemiology (23).
Whole-genome sequencing, de novo assembly, and annotation. High-throughput genome sequencing was carried out on Illumina platforms or a PacBio RS II platform (Novogene, China). The FASTQ reads were quality trimmed with Trimmomatic (v0.36) (24). The draft genome was assembled de novo with SPAdes version 3.0 (25). Antimicrobial resistance genes were identified with ResFinder (26). RAST was used to annotate the sequences of each genome determined with next-generation sequencing (27).
Identification of SNPs, phylogenetic analyses, and definition of clonal group. The core genome of V. parahaemolyticus defined by Gonzalez-Escalona et al. (28) was used as the reference genome to call the SNPs for V. parahaemolyticus (29). SNPs located in the recombination regions were removed by ClonalFrameML (30). Taking 2,000 as the threshold of pairwise SNP distance between strains (14), we defined semiclonal groups (SCGs) in our analyzed data set. Meanwhile, we recalculated the pairwise SNP distance between isolates of each SCG to identify clones which were defined as the genomes with fewer than 10 SNPs differences (14). RAxML version 7.8.6 was used with the generalized time-reversible model and a gamma distribution to model site-specific rate variation (the GTRϩ> substitution model; GTRGAMMA in RAxML) (31). SNPs were recalled for each SCG to gain a higher resolution. ClonalFrameML was used to identify the recombination regions for each SCG. The nonrecombined SNPs were used in GrapeTree (32) to construct the minimum spanning trees. Maximum parsimony algorithms were used for SCG in PAUP 4.0 to precisely identify the SNPs that differed among them (33).
Temporal analysis. Bayesian Evolutionary Analysis by Sampling Trees (BEAST) version 1.8.4 was used to date the important nodes (34). The concatenated nonrecombinant chromosomal SNP alignments of ST415 and ST970 strains were subjected to multiple BEAST analyses with both a strict molecular clock and a relaxed clock, in combination with either constant-size or Bayesian skyline population-size models, to identify the best-fit model, respectively. A marginal likelihood estimation was carried out to obtain path sampling (PS)/stepping-stone sampling (SS) values for each run that had converged in order to compare the different combinations of clock and tree models (35).
Experimental evolution of V. parahaemolyticus and detection of insertion event. To confirm the transferability of ISVal1 from plasmid to chromosome, a coculture experiment of V. parahaemolyticus with ISVal1-containing plasmids was conducted. To prevent the nonspecific amplification of similar insertion sequences in V. parahaemolyticus ATCC 17802, the specific primer pair ISVal1_268F/ISVal1_268R (Table 2) was designed for ISVal1 according to the comparison of the ISVal1 sequence and ATCC 17802 genome sequence. Two consecutive ISVal1 sequences were amplified by PCR from the pirAB-positive plasmid pHZ52-1 and cloned into the SacB-containing plasmid pYC1000-eforRED (36) using Gibson cloning to yield pYC1000-ISVal1. The primers used are shown in Table 2.
Experimental evolution was conducted in a 12-well plate with 2 ml LB30 (LB with 3% NaCl) and 20 g/ml chloramphenicol per well. Every 24 h, 0.02 ml bacterial culture was transferred to a new 12-well plate and incubated at 15°C and 30°C, respectively, under nine parallel experiments that lasted for 10 days. A 20% sucrose-containing LB agar plate was used to eliminate the plasmid. The loss of the plasmid and insertion of ISVal1 were detected in sucrose-resistant and chloramphenicol-sensitive strains by PCR using the primer pairs ATCC-526F/ATCC-526R (an ATCC 17802-specific sequence primer pair to verify that the strain is not contaminated), SacB-1120F/SacB-1120R (SacB gene-specific primers to confirm that the plasmid was eliminated), and ISVal1_268F/ISVal1_268R (an ISVal1-specific primer pair to identify ISVal1 insertion into the genome), which are confirmed by DNA sequencing.
Bacterial growth assay. Triplicates of overnight cultures of V. parahaemolyticus strains were normalized to an optical density at 600 nm (OD 600 ) of 0.1 in TSB30 broth (TSB plus 3% NaCl) and grown at 30°C in TSB30. Growth was measured as the OD 600 . The assay was repeated three times independently, with similar results.
Swimming motility assay. The swimming motility assay was performed on soft LB30 (LB plates containing 0.3% agar) or 1/10 LB30, as described by Yang and Defoirdt (37). V. parahaemolyticus strains were grown overnight in TSB30 broth, and 5-l aliquots (OD 600 , 1.0) were spotted in the center of the soft agar plates. Plates were incubated for 24 h, after which the diameters of the motility halos were measured. All assays were repeated at three times independently. The values are expressed as the mean Ϯ standard deviation (SD).

FIG 6 Legend (Continued)
farming; b, high-frequency genetic exchanges between endemic and oceanic vibrios, resulting in the transfer of plasmid, genomic island, and antibiotic resistance; c, diseased shrimp with a high load of vibrios is shipped to market to minimize the loss; and d, increased human infection cases caused by the consumption of diseased shrimp.

Horizontal Plasmid Transfer
Biofilm formation assay. To compare biofilm formation abilities on abiotic and chitinous surfaces between HZ52 and HZ79, batch experiments were performed in 24-well culture plates. V. parahaemolyticus overnight cultures were incubated at a final concentration of 10 6 cells ml Ϫ1 in 1 ml of marine minimal medium (MMM) (38). Chitin flakes (2% [wt/vol]; Sigma-Aldrich, St. Louis, MO, USA) or 1% (wt/vol) glucose was supplemented as a source of carbon. The microtiter plates were incubated at room temperature with shaking at 60 rpm for 3 days. V. parahaemolyticus biofilm biomass was determined by 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) staining, as described by Sun et al. (39). The experiments were repeated three times. The values are expressed as the mean Ϯ SD.
RNA isolation and quantitative real-time PCR. The VPA1403-1412 (cpsABCDEFGHIJ) operon is responsible for exopolysaccharide production in V. parahaemolyticus, which is controlled by scrABC. To analyze the gene expression of cpsA in HZ52 relative to that of the wild-type strain HZ79, we performed relative quantitative reverse transcription-PCR (qRT-PCR) for cpsA using methods described by Zhang et al. (40). Briefly, overnight cultures of V. parahaemolyticus strains ATCC 17082, HZ52, and HZ79 were subjected to RNA extraction using the TRIzol reagent (Invitrogen, USA). Total bacterial RNA (1 g) was reverse transcribed using qScript cDNA SuperMix (Quanta Biosciences). To measure gene expression, 50 ng of cDNA template from each strain was amplified by real-time qRT-PCR (Applied Biosystems 7900 HT) with Power SYBR green master mix. The qRT-PCR assay was performed on three technical and three biological replicates for each sample. The expression levels of each gene were normalized using an endogenous control gene (DNA gyrase subunit B [gyrB]) to correct for sampling errors. Fold changes in the levels of gene expression relative to ATCC 17082 were measured using the Pfaffl equation (41). The values are expressed as the mean Ϯ SD.
Flow cell and grazing assay. C. hirtus used for the grazing assay was isolated from a shrimp pond and routinely grown on heat-killed Vibrio harveyi strain HW0009 (final concentration, 10 6 cells ml Ϫ1 ). V. parahaemolyticus HZ52 and HZ79 were incubated with or without C. hirtus in a continuous flow cell system, as described by Sun et al. (39). Biofilms of strains HZ52 and HZ79 were grown at room temperature in three-channel (1 by 4 by 40 mm) flow cell tubings, which were attached to a peristaltic pump and continuously fed for 3 days with sterilized marine minimal medium (MMM). One milliliter of MMM containing 10 6 cells ml Ϫ1 of V. parahaemolyticus with or without 10 4 cells ml Ϫ1 of C. hirtus was injected into the cell and allowed to settle for 2 h, after which the medium flow was resumed at a rate of 0.3 ml/min. Chitin flakes were glued to the bottoms of the flow cells using glass silicon. C. hirtus was enumerated daily by inverted microscopy, and the effluents were collected and plated onto TCBS agar to determine the V. parahaemolyticus CFU at day 3. The experiments were repeated three times independently. The values are expressed as the mean Ϯ SD.
Statistical analyses. Data analysis was carried out using the SPSS statistical software (version 18). Bacterial growth data were analyzed by one-way analysis of variance (ANOVA) with a Bonferroni correction. Unless stated otherwise, all other data were compared with independent-samples t tests. A P value of Ͻ0.05 was considered to indicate statistical significance.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.1 MB.