Genomic evidence of environmental and resident Salmonella Senftenberg and Montevideo contamination in the pistachio supply-chain

Pistachios have been implicated in two salmonellosis outbreaks and multiple recalls in the U.S. This study performed an in-depth retrospective data analysis of Salmonella associated with pistachios as well as a storage study to evaluate the survivability of Salmonella on inoculated inshell pistachios to further understand the genetics and microbiological dynamics of this commodity-pathogen pair. The retrospective data analysis on isolates associated with pistachios was performed utilizing short-read and long-read sequencing technologies. The sequence data were analyzed using two methods: the FDA’s Center for Food Safety and Applied Nutrition Single Nucleotide Polymorphism (SNP) analysis and Whole Genome Multilocus Sequence Typing (wgMLST). The year-long storage study evaluated the survival of five strains of Salmonella on pistachios stored at 25 °C at 35% and 54% relative humidity (RH). Our results demonstrate: i) evidence of persistent Salmonella Senftenberg and Salmonella Montevideo strains in pistachio environments, some of which may be due to clonal resident strains and some of which may be due to preharvest contamination; ii) presence of the Copper Homeostasis and Silver Resistance Island (CHASRI) in Salmonella Senftenberg and Montevideo strains in the pistachio supply chain; and iii) the use of metagenomic analysis is a novel tool for determining the composition of serovar survival in a cocktail inoculated storage study.


Introduction
and 89 (Montevideo) isolates from the United States and 2 (Senftenberg) and 7 (Montevideo) 145 from other countries (including Canada, Mexico, and the United Kingdom). The 2016 outbreak 146 isolates were contained in SNP clusters PDS000031814 (Senftenberg) and PDS000027237 147 (Montevideo). There were a total of thirteen pistachio isolates (8 Senftenberg and 5 Montevideo) 148 and 10 clinical isolates (2 Senftenberg and 8 Montevideo) associated with this outbreak that were 149 sequenced and deposited in NCBI. The complete list of isolates and accompanying metadata can 150 be found in Table 1. The isolates were sequenced using the Illumina sequencing chemistry by 151 participating GenomeTrakr laboratories (23), CDC PulseNet laboratories, or private labs. Raw 152 reads for all isolates were downloaded from the Sequence Read Archive (SRA) 153 (https://www.ncbi.nlm.nih.gov/sra) using command line tools from the SRA toolkit available on 154 NCBI. De novo assemblies were generated with SPAdes v3.13.0 (24) using k-mer lengths of [21, 155 33, 55, 77, 99, 127], and the options "--careful" and "--only-assembler".

157
In silico multi-locus sequence typing (MLST) 158 Initial sequence analysis of the isolates was performed using an in silico MLST approach, based 159 on information available at the Salmonella enterica MLST database (25). Seven loci (aroC, 160 dnaN, hemD, hisD, purE, sucA, and thrA) were used for MLST analysis to generate a sequence 161 type (ST). pistachios were completely closed using long read sequencing technology for this study (Table   166 2). Six of the Salmonella isolates were sequenced with the PacBio Sequel system (Pacific Biosciences, Menlo Park, CA, USA) and four isolates were sequenced with the Oxford Nanopore 168 GridION (Oxford Nanopore Technologies, Oxford, UK). The ten closed genomes were 169 submitted in DDBJ/EMBL/GenBank, and annotated using the NCBI Prokaryotic Genome 170 Annotation Pipeline (PGAP) (26). Eight of the complete genomes were reported with a detailed 171 method description in two separate genome announcements (27,28). 172 A brief description of the methods for Pacbio sequencing is as follows. Genomic DNA was   The reads were assembled following previously reported analysis workflows (29). Briefly, the 186 nanopore-generated raw reads were de novo assembled using Canu v1.7 (30). A second polished 187 assembly was generated using SPAdes hybrid assembly (24) (with default settings) using both 188 the Nanopore-generated raw data and the Miseq-generated raw data for each isolate. The two  Gene-by-gene comparison approach 194 To provide gene-by-gene comparison, we aligned all of our closed genomes based on serovar 195 using MAUVE aligner version 20150226 using the progressive algorithm with default settings, 196 and phages were identified using PHASTER (32). Any unique genomic regions, phages, and 197 plasmids identified were imported into Ridom SeqSphere + v. 6.0.2 (Ridom GmbH, Munster, 198 Germany) (33) and used as a reference to interrogate the 201 short read isolate assemblies used 199 in this study for presence or absence. software v 6.0.2 (33) by creating ad hoc wgMLST databases derived from the closed genomes 205 previously described. Using the cgMLST target definer tool and accessing our genomes from 206 NCBI, an ad hoc wgMLST scheme for Salmonella Senftenberg was developed. The scheme was 207 based on the closed reference genome CFSAN047866 (CP029040), comprised of 4,768 genes, 208 and query genomes (CP029036, CP029038, CP037892, and CP037894) found in Table 2 using 209 the following default thresholds. For the reference genome filter thresholds: (1) minimum length: (3,908 core targets and 640 accessory targets). The 106 Salmonella Senftenberg assemblies (54 216 nuts/seeds, 44 environmental, 7 clinical and 1 poultry) were typed using this scheme. The alleles 217 identified were used to establish a matrix disregarding missing allele values. A minimum 218 spanning tree was constructed to visualize the results using the Ridom SeqSphere + software. 219 Clusters were automatically assigned using the default distant threshold for the software of ≤ 10 220 alleles.

221
The same process was repeated to create an ad hoc wgMLST scheme for Salmonella Montevideo 222 using the closed genome CFSAN051296 (CP029336), comprised of 4,664 genes, as the 223 reference and the four other closed genomes from this study (CP029035, CP029039, CP040379, 224 and CP040380) as query genomes using the previously defined settings to define the core 225 genome loci. The resulting scheme consists of 4,447 loci (3,876 core targets and 571 accessory 226 targets). The 95 Salmonella Montevideo assemblies (49 nuts/seeds, 30 clinical, 9 environmental, 227 2 poultry, 2 missing source, 1 beef, 1 swine, and 1 cattle) were typed using this scheme. A 228 minimum spanning tree of all isolates was constructed using the software from the allele calls 229 disregarding missing values, and clusters were defined with the same threshold as above. Senftenberg) associated with low moisture food commodities and one strain of Salmonella 250 Newport isolated from tomatoes was used in this experiment (Table 3). The genomes of all five 251 isolates were completely closed using Pacbio long read technology (27,36). For this study, 252 pistachios were inoculated with the 5-strain cocktail, as well as inoculated with the same strains 253 on an individual basis. Inocula were prepared from stock culture as described in Keller et al. (37) 254 with slight modifications. Briefly, a single colony from each working stock was transferred to 255 Trypticase Soy Broth (TSB, BD, Franklin Lakes, NJ, USA) and incubated overnight at 37°C.

256
After incubation, 100 µl of inoculum was spread plated on Trypticase Soy Agar (TSA, BD) 257 plates. The plates were incubated overnight at 37°C and then bacteria were harvested by adding 258 1.0 ml of Buffered Peptone Water (BPW, BD) to the surface of the plate and gently scraped 259 using an L-shaped plate spreader. Each plate yielded ~0.5 ml of harvested culture at 260 approximately 10 log CFU/ ml. The cocktail was formed by combining equal volumes of all five 261 harvested cultures and used for inoculation of the pistachios. Organic, raw inshell pistachios 262 were purchased and tested for background microbial populations using the Biomerieux Tempo 263 AC kit (Biomerieux, Durham, NC, USA). A high-population cell suspension was prepared by 264 combining 18 ml of the five-serotype Salmonella cocktail with 1,800 ml of sterile deionized 265 water. A total of 900 g of pistachios were soaked in the suspension bath for 1 minute, and then 266 placed on trays lined with absorbent towels in a biosafety cabinet to airdry overnight at room 267 temperature. After 24 h, pistachios were stored in desiccator cabinets equilibrated to either 35% 268 relative humidity (RH) or 54% RH at 25 o C, which were selected to represent a low and high 269 water activity (a w ) within the range of a w levels typical of low moisture foods (38). Uninoculated 270 control pistachios were held under the same conditions. To determine the surviving populations, 271 triplicate (10g) samples were removed for analysis at 0, 1, 2, 4, 6, 13, 27, 55, 83, 180, 270, and 272 365 days. The water activity was recorded for the inoculated and control pistachios from the 273 different relative humidity conditions using an Aqualab model 4TE water meter (Decagon 274 Devices, Pullman, WA). Two random 10 g samples from the uninoculated control group were 275 tested using the Biomerieux Tempo AC kit to determine background microbe counts. The 276 triplicate samples of inoculated pistachios were diluted 1:9 in BPW and appropriate dilutions 277 were plated in duplicate on the differential media m-TSAYE [Trypticase Soy Agar with Yeast 278 Extract supplemented with 0.05% (wt/vol) ammonium iron citrate and 0.03% sodium 279 thiosulfate]. After overnight incubation at 37°C, colonies were counted and log CFU/g was 280 calculated. A repeated measure analysis of variance (ANOVA) was performed in R v3.6.1 to 281 determine the differences in log CFU/g over time between the high and low humidity conditions.

282
The differences in the slopes (or rate of reduction in log CFU/g) between the two humidity 283 conditions were also investigated in R. profile set to 0, and minimum mean percent coverage set at 80%. Since we were looking at the 303 presence or absence of known serovars in our inoculum, the threshold for minimum mean 304 percent coverage was lowered below 90%. This tool utilizes MLST to assign reads to alleles and 305 determines STs, which allowed us to determine the serovar present, since our isolates have 306 distinct STs.

307
The single-strain storage study was conducted in parallel to the 5-strain cocktail storage study.

308
The same procedure was used to prepare a high population cell inoculum of each individual 309 strain and to inoculate 400 g portions of pistachios with each serovar. The same storage 310 conditions were used and triplicate (10 g) samples were removed for analysis at Day 0, 1, 7, 28, 311 56, and 84. The triplicate samples were diluted 1:9 in BPW and appropriate dilutions were plated 312 in duplicate on the m-TSAYE for direct plate counts.  The MLST results for Salmonella Montevideo isolates also detected two STs. SNP cluster 325 PDS000027237 is a member of ST316, while SNP cluster PDS000032600 belongs to ST138.

326
The ST316 group contains a total of 82 isolates, of which 24 were clinical isolates and 33 were 327 isolates from pistachio sources ranging in collection date from 2009 to 2018. ST316 has 328 previously been described as the "outbreak" clade due to its linkage with numerous outbreak-329 associated plant products including tahini, sprouts, spices, as well as pistachios (42). The ST138 330 group of 13 isolates contains 6 clinical isolates and 3 isolates from pistachios. These sequencing 331 typing results can be found in Table 1  representatives of each ST to use as a reference for SNP analysis, as well as to identify the 337 positions of mobile elements and assess genetic differences between STs. These specific isolates 338 were selected as they were sourced from pistachios, readily available in our laboratory, and had 339 different collection dates. In this study, we used two different types of long read sequencing 340 technology, Pacific Biosciences (PacBio) and Oxford Nanopore Technology (ONT), to generate 341 the closed genomes (27, 28). The S. Senftenberg isolates ranged from 4.7 -4.8 mb in size while 342 the S. Montevideo isolates were ~ 4.6 mb. Those isolates sequenced with PacBio had genome 343 coverage from 204 -500×, and those sequenced with ONT had 25 -103× coverage. One plasmid, 344 over 300 kb in size, was found associated with Salmonella Senftenberg CFSAN000878. The 345 genome statistics for these ten isolates can be found in Table 2.

347
Genome-wide comparisons 348 The content of the ten closed genomes of Salmonella consisting of 2-3 representatives from each 349 sequence type were compared to each other, since many NGS analysis algorithms mask mobile 350 elements such as phages, plasmids, and pathogenicity islands and which may lead to important 351 characteristics of an isolate being overlooked. By comparing the complete and closed genomes 352 from different isolates, we can find and identify genomic elements that a clonal strain may 353 acquire or dispose of over time. The closed genomes were evaluated for conserved genes shared 354 within the same sequence type, as well as between both serovars (Senftenberg and Montevideo), 355 using the Mauve genome aligner, PHASTER, and BLAST. As expected, the isolates within the 356 same sequence type showed higher similarity.

357
Interestingly, one unique region was identified that was found to be shared between the ST14 358 and ST316 isolates but missing in the ST185 and ST138 isolates. This unique region was 359 determined to be the Copper Homeostasis and Silver Resistance Island (CHASRI) and is about 360 21 kb in size (43). Based on analysis using Ridom SeqSphere + , this island was confirmed in all 361 of the short read sequence data of isolates associated with ST14 (N=54) and ST316 (N=82). The 362 island is located on the chromosome in both serovars and contains the genes pcoABCDEFGRS, 363 cusABCFRS, silE, and silP. The sequence of the CHASRI had a 100% identity intra-serovar and 364 a 99.95% identity inter-serovar when the region was analyzed with BLAST (44). 365 The CHASRI has been identified previously in Escherichia coli, Enterobacter cloacae, 366 Klebsiella pneumoniae, and Salmonella enterica (43). This island has previously been shown to 367 increase the ability of a bacterium to survive in an environment with higher levels of copper and 368 silver, as well as the ability to provide tolerance during the transition between aerobic and 369 anaerobic environments, affording a fitness advantage for facultative anaerobes (43, 45). The use 370 of copper as a growth promoter in swine and poultry has led to the selection of this island in 371 Salmonella associated with these sources in other countries (46,47). The use of copper as a 372 fungicide and bactericide in agriculture has been reported for many years, and it is generally 373 considered safe for both conventional and organic produce use (48,49). Copper deficiency is 374 reportedly common for pistachios and the use of copper in foliar and soil treatments has been 375 recommended to the industry (50). Copper-containing foliar sprays may be applied at various 376 times while pistachios are developing in the trees (51). The presence of the CHASRI in the 377 pistachio-associated isolates may be an acquired adaptation by these specific salmonellae in 378 response to copper stress in the growing environment.

379
The IncH12/IncH12A plasmid (CP029037) identified in Salmonella Senftenberg 380 CFSAN000878, with a length of 319,930 bp, contains multiple heavy metal resistance genes 381 which confer arsenic resistance (arsB, arsC, arsH), mercury resistance (merA, merC, merD, 382 merP, merR, merT), and tellurium resistance (terABCDEF), as well as a copy of silE, cusA/czcA, 383 and cusS. The plasmid also contains one antimicrobial resistance gene, MCR-9.1(52, 53), which 384 confers Colistin resistance and is of significant public health importance. The plasmid sequences 385 for the duplicated genes, silE, cusA/czcA, and cusS, differ from the same genes found in the 386 chromosome with 91.2%, 93.7%, and 95.1% identity, respectively. This plasmid was only 387 identified in one other isolate (CFSAN016599) from pistachio also collected in 2009. Since the 388 plasmid has not been identified in any isolates after 2009, it is very likely that it was lost over 389 time due to its large size and the genes on this plasmid not being necessary but rather redundant 390 for increased fitness due to the incorporated CHASRI already in the chromosome.

391
Phages can encode factors that increase the virulence of Salmonella by enhancing adhesion, 392 intracellular survival, and host entry (54). Phages were identified using PHASTER, and those 393 sequences were then used as references in Ridom SeqSphere + to interrogate the short-read 394 assemblies. One intact phage, Salmon_SPN3UB_NC019545, was found in one of the closed 395 genomes from ST14. Specifically, this phage was found in the closed genome of CFSAN047866 396 (CP029040) as well as in the short read data from 7 of 54 isolates by analysis with Ridom SeqSphere + . Interestingly, these isolates were not associated with a single group on the SNP tree.

398
The two closed isolates from ST185 contain two phages: Aeromo_phiO18P_NC009542 and 399 Salmon_118970_sal3_NC031940. Using Ridom SeqSphere + to analyze the short-read data, the 400 first phage was found in 52 of 52 isolates, and the second phage was found in 5 of 52 isolates cgMLST approach is the ability to analyze isolates without the need for a reference sequence. In 415 this study, two ad hoc wgMLST schemes were created for analysis of the S. Senftenberg and S. 416 Montevideo isolates.

417
The Salmonella Senftenberg wgMLST scheme of 4,548 loci (3,908 core targets and 640 418 accessory targets) that was produced using the five closed Salmonella Senftenberg genomes 419 isolated from pistachios was used to analyze the assemblies of the 106 Senftenberg isolates 420 selected for our study. A minimum spanning tree produced by analysis of all common allele calls 421 from both the core genome and accessory genome shows that Senftenberg isolates in ST14 are 422 distinct from isolates in ST185 by a difference of 3,534 alleles ( Fig. 1). This allele difference 423 between ST14 and ST185 is meaningful and illustrates that the primary contamination sources 424 for these isolates are not the same. The analysis separated the isolates belonging to ST14 into 425 three clusters (Clusters 1, 2, and 3) and the ST185 isolates into one cluster (Cluster 4). The 426 pistachio isolates from ST14 have more branching and allele difference (range of 0-11) over an 427 eight-year time period (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017). The majority of the isolates belong to Cluster 1, which 428 contains isolates associated with the 2016 outbreak. One pistachio isolate, CFSAN015119, from 429 2009 located in Cluster 2 had a higher allele difference (≥10) from the other pistachio isolates.

430
Cluster 3 is located 20 alleles away from the nearest pistachio isolate and is associated with 431 isolates from chicken. The pistachio-associated isolates from ST185 have low allele differences 432 (0-7) over a five-year time period (2013-2018), which illustrates genome conservation and 433 clonality. It is noteworthy that a distinct central node was found with the ST185 isolates. Finally, 434 the isolate FSE0085, isolated from contaminated tahini, has 50 allele differences from the central 435 node and is not related to Cluster 4.

436
The wgMLST scheme (3,876 core loci and 571 accessory loci) that was created using the five 437 closed Salmonella Montevideo genomes isolated from pistachios was used to analyze the 95 438 Salmonella Montevideo isolates from our study. A minimum spanning tree produced by distance 439 calculation for all common allele calls from both the core genome and accessory genome was

457
There are two clinicals, PNUSAS041766 and PNUSAS045784, that are a match to each other 458 with our wgMLST scheme and are one allele from the large node of outbreak-associated isolates.

459
Also, three clinical isolates with 3-4 allele differences from an ST138 pistachio isolate are in 460 Cluster 3. One of these isolates, 12-1128, is from Canada. Unfortunately in this case, without 461 epidemiological data and more complete metadata from these isolates, it is impossible to confirm 462 association with outbreak strains or the consumption of contaminated pistachios.

463
The results of our wgMLST analysis also showed the presence of large differences between 464 isolates within the same serovar but belonging to different STs. For Salmonella Senftenberg,we 465 were able to visualize the genomic separation of the two ST clusters, witness multiple mutation 466 events in the ST14 strains, and see clonality within the ST185 isolates. The results from 467 Salmonella Montevideo also showed the delineation of isolates from the different STs. We also 468 identified multiple clinical isolates from the UK, Canada, and US that were not previously 469 associated with an outbreak but were closely related to previous pistachio outbreak isolates from finding that all of these isolates contain the CHASRI islandcoinciding with the known use of 512 copper in pistachio orchards providing selective pressure for this adaptation, as well as previous 513 studies showing poultry isolates containing this islandfurther supports that these strains have 514 established residence in or around the production environment. Although, the initial source of the 515 contamination (e.g., adjacent animal operations) of this strain can not be determined, our analysis 516 shows that it was able to establish residence.

517
For the second Senftenberg tree, closed genome CFSAN010508 was used as the reference pistachios or equipment to be shared or exchanged between the two facilities. It seems unlikely 534 that two strains with no SNP differences would occur in disparate facilities without some type of 535 indirect or direct relationship between them. Rather, the small number of genetic differences 536 over an extended period of time would suggest the transmission of the same strain between the 537 two facilities and therefore a clonal strain taking harborage (20, 67). Further, the lack of any 538 measurable SNP distance between these Salmonella isolates from these two facilities suggests 539 that such residence would be in an environment with relatively low environmental stress, such as 540 an environment devoid of implementation of more robust sanitation controls.

568
A phylogenetic tree was also constructed for the 13 Montevideo isolates from ST138 (Fig. 4b). A 569 total of 48 distinct SNP variants were identified by reference-based SNP analysis in the isolates 570 of ST138 using CFSAN010508 as the reference genome. Clinical isolates are highlighted with a 571 star on the tree (n=6). This sequence type has been reported to be primarily associated with 572 bovine sources (25), and indeed, two bovine isolates, PNSUAS004495 and FSIS31800836, were 573 found to be distantly related by 30 -47 SNPs from the pistachio isolates. There are three 574 pistachio isolates from three different facilities (Facility A, Facility G, and Facility I). The cluster 575 that includes the three pistachio isolates shows high relatedness with 1 -4 SNPs (median 3 576 SNPs). Three isolates from Facility G, two environmental swabs and one pistachio, also belong 577 to this cluster. The SNP distance between the strain isolated in 2009 and the 2015 isolate and 578 2017 isolate is only 2 SNPs. The isolates of this cluster, both pistachios and swabs from three 579 different facilities (Facility A, Facility G and Facility I), were collected over a wide timespan (8 580 years) and have low diversity, which suggests a persistent Salmonella strain. This strain is likely 581 to reside in a lower-stress environment than in the orchard environment, which could include 582 shared farm equipment, transport vehicles, or an intermediate space. The low SNP distances 583 suggest a strain that is not undergoing selective pressure and may also have a slower generation 584 time similar to a resident pathogen (20). There are also three clinical isolates that are within 1 -585 8 SNPs (median 3 SNPs) from the pistachio isolates. However, without epidemiological data we 586 are unable to positively link these illnesses with contaminated pistachio exposure.

588
Storage study to assess persistence in pistachios 589 In order to further explore the nature and likelihood of long-term environmental persistence of 590 these Salmonella, we tested the persistence of Salmonella strains both individually and in a 591 cocktail after inoculation on raw inshell pistachios. The strains used in this study were associated 592 with low-moisture food products (Anatum-peanuts, Oranienburg-pecans, Montevideo-pistachios, 593 Senftenberg-pistachios) and a strain isolated from tomato (Newport) ( Table 3) (Table 3). The basis behind using the Newport strain was to 598 determine if there were differences in survivability on a low-moisture food based on the initial 599 source of the isolate (i.e., a high-moisture food). The Salmonella in the five-strain cocktail was 600 able to persist in pistachios for up to one year (the length of the study) (Fig. 5). There was a 1 601 log CFU/g reduction in Salmonella during the period of desiccation (24 hour drying time) prior 602 to storage. The Salmonella in the inoculated pistachios stored at 35% relative humidity (mean a w 603 = 0.3398, stdev = 0.0228) were able to persist for a year with only an additional 1 log CFU/g reduction after desiccation (a total of 2 log CFU/g reduction). The Salmonella in the pistachios 605 stored at 54% relative humidity (mean a w = 0.5337, stdev = 0.0370) showed an additional 3 log 606 CFU/g reduction over a year, and a repeated measures ANOVA analysis showed the impact of 607 humidity condition (high vs low) over time. The rate of reduction was higher for the high 608 humidity condition based on the observed difference in the slope (rate of reduction in log CFU/g 609 for low and high humidity was -0.124 and -0.264 per day, respectively) between the two 610 conditions (p<.0001). Differences in log CFU/g between low and high humidity conditions was 611 most pronounced after 6 months (Fig. 5). These findings are consistent with those reported in 612 literature (37, 38).

613
One of our study aims was to determine if there was a difference in which serovars were 614 surviving within the cocktail inoculum. A previous storage study by Kimber et al. (72)  noted where colonies from a single serovar were not selected for the molecular serotyping assay 624 but was able to be detected by metagenomic sequencing. Sequencing was conducted post-625 enrichment, therefore the serovars identified from the reads is from viable bacteria. The 626 limitation of the molecular serotyping assay is the number of colonies that are selected and 627 tested. Unintentional bias could be introduced into the results based on colony selection by either 628 over-or under-selection, but there is a higher probability of sequencing any and all strains that 629 are present above a certain threshold with the sequencing method. The combined serotyping 630 results from both methods showed that all five serotypes could be detected at one year of storage.

631
The results from the metagenomic ST assignment with SRST2 (Fig. 6b) concurred with the 632 Luminex assay (Fig. 6a) results most of the time, but due to the limitation of sequencing depth, 633 Oranienberg was not identified in several instances. The sequencing runs consisted of 634 multiplexing 8-10 samples per run. However, this level of multiplexing did not provide enough 635 sequencing depth to detect Oranienberg, which may have been occurring at lower concentrations.  Each individual strain was also used to inoculate pistachios and stored at the same low and high 649 relative humidity to compare the results of persistence of a single strain contamination versus a 650 multi-strain contamination event. No significant differences in the bacterial counts between 651 strains or between storage conditions (data not shown) were found during our three-month 652 storage period, and therefore, these trials were halted at Day 84.

653
The results from our study conclude that the five-strain cocktail of Salmonella can survive in Pistachios may be contaminated with pathogenic bacteria in the growing, production, and 666 processing environments and present a hazard due to the ability for Salmonella to persist in low-667 moisture environments and foods. Diagnostic capabilities have been greatly enhanced due to the 668 development and increasing applicability of WGS, which is now fully deployed as a molecular 669 epidemiological tool to assist in foodborne outbreak investigations (17)(18)(19). Epidemiological data 670 are an important complement to WGS for affirming exposure to a particular pathogen.

671
Nonetheless, by using WGS in two different phylogenetic analyses, we were able to match 3) a clonal strain at the pistachio handling facilities, which could include storage silos.

685
Our storage study demonstrates that Salmonella will survive in pistachios for a minimum of one 686 year at both high and low relative humidities. The use of a non-nut associated strain, Salmonella 687 Newport, in our cocktail and its ability to be recovered at the 12 month time point, showed that