Salmonella enterica subspecies enterica is traditionally subdivided into serovars by serological and nutritional characteristics. We used Multilocus Sequence Typing (MLST) to assign 4,257 isolates from 554 serovars to 1092 sequence types (STs). The majority of the isolates and many STs were grouped into 138 genetically closely related clusters called eBurstGroups (eBGs). Many eBGs correspond to a serovar, for example most Typhimurium are in eBG1 and most Enteritidis are in eBG4, but many eBGs contained more than one serovar. Furthermore, most serovars were polyphyletic and are distributed across multiple unrelated eBGs. Thus, serovar designations confounded genetically unrelated isolates and failed to recognize natural evolutionary groupings. An inability of serotyping to correctly group isolates was most apparent for Paratyphi B and its variant Java. Most Paratyphi B were included within a sub-cluster of STs belonging to eBG5, which also encompasses a separate sub-cluster of Java STs. However, diphasic Java variants were also found in two other eBGs and monophasic Java variants were in four other eBGs or STs, one of which is in subspecies salamae and a second of which includes isolates assigned to Enteritidis, Dublin and monophasic Paratyphi B. Similarly, Choleraesuis was found in eBG6 and is closely related to Paratyphi C, which is in eBG20. However, Choleraesuis var. Decatur consists of isolates from seven other, unrelated eBGs or STs. The serological assignment of these Decatur isolates to Choleraesuis likely reflects lateral gene transfer of flagellar genes between unrelated bacteria plus purifying selection. By confounding multiple evolutionary groups, serotyping can be misleading about the disease potential of S. enterica. Unlike serotyping, MLST recognizes evolutionary groupings and we recommend that Salmonella classification by serotyping should be replaced by MLST or its equivalents. Microbiologists have used serological and nutritional characteristics to subdivide pathogenic bacteria for nearly 100 years. These subdivisions in Salmonella enterica are called serovars, some of which are thought to be associated with particular diseases and epidemiology. We used MultiLocus Sequence-based Typing (MLST) to identify clusters of S. enterica isolates that are related by evolutionary descent. Some clusters correspond to serovars on a one to one basis. But many clusters include multiple serovars, which is of public health significance, and most serovars span multiple, unrelated clusters. Despite its broad usage, serological typing of S. enterica has resulted in confusing systematics, with a few exceptions. We recommend that serotyping for strain discrimination of S. enterica be replaced by a DNA-based method, such as MLST. Serotyping and other non-sequence based typing methods are routinely used for detecting outbreaks and to support public health responses. Moving away from these methods will require a major shift in thinking by public health microbiology laboratories as well as national and international agencies. However, a transition to the routine use of MLST, supplemented where appropriate by even more discriminatory sequence-based typing methods based on entire genomes, will provide a clearer picture of long-term transmission routes of Salmonella, facilitate data transfer and support global control measures.

Very high-throughput sequencing technologies need to be matched by high-throughput functional studies if we are to make full use of the current explosion in genome sequences. We have generated a very large bacterial mutant pool, consisting of an estimated 1.1 million transposon mutants and we have used genomic DNA from this mutant pool, and Illumina nucleotide sequencing to prime from the transposon and sequence into the adjacent target DNA. With this method, which we have called TraDIS (transposon directed insertion-site sequencing), we have been able to map 370,000 unique transposon insertion sites to the Salmonella enterica serovar Typhi chromosome. The unprecedented density and resolution of mapped insertion sites, an average of one every 13 base pairs, has allowed us to assay simultaneously every gene in the genome for essentiality and generate a genome-wide list of candidate essential genes. In addition, the semiquantitative nature of the assay allowed us to identify genes that are advantageous and those that are disadvantageous for growth under standard laboratory conditions. Comparison of the mutant pool following growth in the presence or absence of ox bile enabled every gene to be assayed for its contribution toward bile tolerance, a trait required of any enteric bacterium and for carriage of S. Typhi in the gall bladder. This screen validated our hypothesis that we can simultaneously assay every gene in the genome to identify niche-specific essential genes.