Genome divergence and increased virulence of outbreak associated Salmonella enterica subspecies enterica serovar Heidelberg

In the summer of 2016, several Midwestern premises raising bottle-fed calves experienced striking losses in dairy bulls purchased from sale barns. Death rates, usually dictated by neonatal diarrhea which is very common in sale barn calves, increased 5–10-fold. Illness was subsequently found in dairy heifers as well. Disease was apparent even when best practices were used: it affected quality calves given space and comfort, on a good plane of nutrition, with no failure of passive transfer. Sporadic cases were also found dating back to 2014 in our database (see Table 1).

Predominantly young, < 3-weeks-old dairy calves were the source of S. Heidelberg. Pre-existing or concurrent disease was found in 9 of 12 younger calves and 3 of 4 older calves, but detection of Salmonella ser. Heidelberg correlated with markedly increased death losses clinically, comparable to those seen in herds infected with S. Dublin, a known serious pathogen of cattle. Salmonellosis was suspected at autopsy or by histopathologic examination in 5 of 12 young calves, and in 0 of 4 older calves, highlighting the importance of bacterial isolation. Since no lesions of salmonellosis were found in older calves, a carrier state cannot be ruled out.

To determine the possible genomic changes that would have contributed to the increased virulence and its relationship to the strains causing human infections, a reference based SNP analysis of 634 Salmonella ser. Heidelberg strains was performed with Salmonella ser. Heidelberg str. SL476 (NCBI accession: NC_011083.1) as a reference genome. This whole genome sequence based SNP comparison revealed 1025 variant positions across the study isolates. Phylogenetic arrangement of genomes in the increasing order of branch length showed clusters of genetically related strains (Fig. 1). Based on our analysis, a chicken isolate from the USA (USA242) with lesser variants and branch length was positioned phylogenetically higher from all other isolates followed by isolate USA488, indicating chicken as the possible evolutionary origin of Salmonella ser. Heidelberg strains in this outbreak. Recently sequenced wild avian strains by our group (USA 481, USA482, and USA 483), collected in the early 90s, were positioned in the top few clusters closely related to chicken and human isolates.

As the branch length increased, we identified several major clusters in the SNP based phylogenetic analysis of our study isolates (Fig. 1). With increasing branch length, more clusters were differentiated mainly based on isolation sources. An exception was the cluster formed by around 80% of the isolates from the United Kingdom, which contained 25 SNPs that were unique to this cluster. Another interesting observation was that most of the egg isolates remained as a distinct clade having 138 SNPs unique to this cluster. Further down in the SNP tree, more than 50% of isolates from pigs formed a cluster which might be an indication of the expanding host colonization ability of Salmonella ser. Heidelberg strains. Two of the most divergent clusters, located at the bottom of Fig. 1 in our analysis, primarily originated from cattle, turkey and human cases. The bovine sub-cluster in this analysis was composed of 32 isolates and most were clonal. Defining features of these isolates is given in Table 1. Strain USA460, a bovine isolate, formed an outlier in this cluster indicating bovine origin of all isolates in this cluster. The bovine strains isolated from the animals believed to be originating from Wisconsin closely clustered with human and food isolates. For example, strains USA538 and USA582 originating from ground beef clustered closely with strains USA104 and USA107, which were isolated from human blood. Strains USA114, USA115, USA116 and USA117, environmental isolates from a livestock sale barn in Wisconsin, clustered in the middle of a clonal group of strains that were of human and bovine origin. The outgroup of this cluster was composed of strains that were isolated in the 1980s (USA 141, chicken isolate, 1987; USA128, porcine isolate, 1987) to recent years (Canada008, human isolate, 2012; Canada006, human isolate, 2012) and also from diverse hosts. Therefore, it appears that the strains related to the outbreak associated strains were present in a wider geographical region and a number of hosts other than cattle.

To understand the virulence repertoire of Salmonella ser. Heidelberg isolates, we carried out a virulence gene mapping of all 634 genomes. To predict the presence of virulence genes, we performed a local BLAST search of the genomes against virulence gene sequences obtained from the virulence factor database (VFDB) [28, 29]. We grouped 184 identified salmonella virulence genes into four groups (Fig. 2). As expected, the core set of Salmonella virulence genes were conserved among all isolates irrespective of isolation source and their geographical origin. However, isolates in the bovine cluster (Table 1) contained all the 4 fimbrial adherence determinants (safA, safB, safC and safD) of the Salmonella atypical fimbriae (Saf) gene cluster, while the full cluster was absent in all other isolates compared in this study. A detailed result of the virulence gene mapping across all genomes is given in Additional file 2: Tables 2A, B. Therefore, the virulence mapping of isolates revealed the main hallmark of outbreak associated strains was the presence of safABCD operon genes which were absent in all other isolates compared in this study.

To determine the antibiotic resistance genotype among the Salmonella ser. Heidelberg isolates, we used a comparative genomic analysis approach. We performed a local BLAST search of all 634 of Salmonella ser. Heidelberg genomes against > 2000 antimicrobial resistance genes in the ResFinder database. A full list of resistance genes identified among these genomes with a minimum 95% sequence identity is given in Additional file 3: Tables 3A, B. Our results show that 50% of the study isolates (317/634) contained least one resistant gene (Fig. 3). Among isolates from chicken, 37% (75/202) were resistant while resistance was 83% (100/120) among turkey isolates. Highest resistance rate (93%) was identified among isolates of bovine origin. Among isolates of human origin, 30% percentage (47/154) contained at least one antibiotic resistance gene. One interesting observation was that, at 95% sequence identity, we could not predict any quinolone resistance genes in poultry isolates except in one turkey isolate (USA525) which was identified with genes qnrS2 and qnrS6. Comparison dataset in this analysis contained isolates dating back to 1980’s. When the resistance profiles of older isolates were compared to recent outbreak causing isolates, we find that the newer isolates contained higher number and type of resistance genes indicating acquisition of more antibiotic resistance among Salmonella ser. Heidelberg over these years.

Seventeen Salmonella ser. Heidelberg strains were isolated from cases submitted to the Animal Disease Research and Diagnostic Laboratory, South Dakota State University (ADRDL-SDSU) and Wisconsin Veterinary Diagnostic Laboratory (WVDL) per standard operating procedures. Pathological investigation was carried out following the guidelines set by the American Association of Veterinary Laboratory Diagnosticians. After Salmonella isolation and identification, strains were cultured overnight in Luria–Bertani broth. DNA was extracted from 1 mL of the overnight culture using Qiagen’s DNeasy blood and tissue kit (Qiagen, Inc., Valencia, CA) according to the manufacturer’s protocol. Quality of DNA was assessed using Nanodrop One™ (Thermo scientific™, DE). Quantity of DNA was measured using Qubit^® 3.0 (Thermo Fisher Scientific Inc., MA) fluorometer and stored at − 20 °C until further use.

Sequencing library was prepared using Nextera XT kit (Illumina Inc. San Diego, CA). Concentration of the DNA samples were adjusted to 0.3 ng/µl prior to processing. After bead normalization, DNA libraries were pooled at equi-molar concentration. Pooled libraries were then sequenced on Illumina Miseq platform (Illumina Inc., CA) using V2 chemistry to generate 2× 250 paired end reads.

The raw data files for the seventeen isolates sequenced by our group were de-multiplexed and converted to FASTQ files using Casava v.1.8.2. (Illumina, Inc, San Diego, CA). For comparative analysis, we downloaded publicly available genome sequence data in Fastq format for 617 Salmonella ser. Heidelberg isolates from the National Center for Biotechnology Information Sequence Read Archive (NCBI-SRA) database using NCBI-SRA tool kit (version 2.8.1-2). To ensure a uniform analytic workflow, we considered only genomes sequenced using the Illumina platform. This constituted 634 genomes in total. Metadata for all genomes are given in Additional file 1: Table S1. Genome data was then imported into CLC Genomics work bench (version 9.5.3, Qiagen bioinformatics) and was further processed. Failed reads as well as reads having length below 200 and above 500 bases were removed. A quality based trimming, with a quality score limit of 0.05 as described by CLC Genomics workbench, was performed to ensure good quality sequence data without any ambiguous nucleotides. Trimmed reads were used for SNP based phylogenetic analysis while assembled genomes were used for virulence gene mapping and antimicrobial resistance gene profiling. For SNP analysis, we used quality trimmed reads mapped to reference genome Salmonella ser. Heidelberg str. SL476 (NCBI accession: NC_011083.1). This genome was selected as reference as it is a complete genome and has been used as reference for comparative genomics studies of Salmonella ser. Heidelberg previously [35‐38]. For variant calls, fixed ploidy variant detection for bacterial genomes with a minimum 90% variant probability was used. Variant calls and read mappings results were then used to determine the SNP positions and to estimate the consensus sequence. Neighborhood joining method was used to create the phylogenetic tree composed of all 634 genomes. For virulence and antimicrobial gene mapping, we used local databases of Salmonella enterica virulence gene sequences available from Virulence Factor Database (VFDB) [28, 29] and ResFinder [39], respectively. Genome assemblies were then used for BLAST searching against these local gene databases with criteria of ≥ 95% minimum sequence identity and ≥ 50% minimum sequence length.