Early MLST: seven genes, eBURST and the legacy of sequence types
Ed Feil and Natacha Couto join the hosts to discuss the origins and continuing use of multi-locus sequence typing (MLST). They explain how seven-gene schemes emerged, why digital sequence data changed bacterial epidemiology, and how eBURST offered an alternative to dendrograms. The conversation also examines recombination, weaknesses in gene selection and the lasting value of sequence-type names.
Listen to Episode 102: Early days of MLST on SoundCloud
In this episode
Why seven genes?
Ed traces MLST back to multi-locus enzyme electrophoresis (MLEE), which measured enzyme mobility in starch gels. Comparing samples required running them alongside known strains, making the technique difficult to transfer between laboratories. The established Neisseria scheme used about 15 loci, and the new sequence-based method needed to recover its recognised lineages.
Ed dates the original MLST paper to 1998 and explains that it actually used 11 genes: six housekeeping genes and five more variable genes, including outer-membrane proteins. The variable genes added little to the grouping. Six housekeeping genes nearly reproduced the MLEE lineages; a seventh separated a remaining group. Seven was therefore a practical balance between cost and resolution, not a universal biological requirement.
From gene fragments to sequence types
MLST compares partial sequences of up to about 500 base pairs from a small set of housekeeping genes. Ed explains that fragment length reflected what sequencing runs in opposite directions could reliably recover. Genes were chosen to be evolutionarily unremarkable and under purifying selection, aiming to represent underlying relatedness rather than rapidly changing traits.
Each distinct sequence at a locus receives an allele number. The resulting profile resembles a telephone number, and each unique profile receives a sequence type, or ST. Alleles differing by one base or many bases are both counted as different: recombination can introduce several changes in one event. A valid ST requires every locus. Crucially, digital sequences allowed laboratories worldwide to compare their results through internet-accessible databases.
MLST versus PFGE
Natacha recalls starting her PhD in 2011 without access to whole-genome sequencing. MLST provided a shared language for discussing lineages, but she also used pulsed-field gel electrophoresis (PFGE) when investigating similarities between bacterial strains from animals and humans. Finding ST398 in both was not enough resolution for her transmission questions.
Ed describes PFGE as a restriction-fragment, band-based method whose genome-wide reach could detect accessory-genome variation missed by MLST. Its relative discriminatory power depended on the organism. However, PFGE required dedicated laboratory expertise, and scanned gel patterns were less straightforward to store and compare than sequence data. Natacha describes PFGE’s role in MRSA outbreak investigations, alongside alternatives such as MLVA and rep-PCR.
Why Ed developed eBURST
Early MLST results were commonly displayed as UPGMA dendrograms. Ed remembers his supervisor using a magnifying glass to read their labels, even for datasets of roughly 100 strains. He wanted a display that better represented the apparent process of clonal expansion.
In Staphylococcus aureus and Neisseria meningitidis data, Ed saw central genotypes surrounded by variants differing at one locus. eBURST placed the inferred founder centrally, with related profiles around it, rather than showing ancestors and descendants at the same level in a tree. The discussion then turns to minimum spanning trees: many equally minimal connections are possible, so prioritising genotypes with the most near neighbours supplies an additional model for choosing a solution.
Imperfect schemes and recombining populations
Natacha questions whether schemes for organisms such as Pseudomonas and Acinetobacter selected the best genes. Ed agrees that early choices were made before genome sequences were available and could have been improved. He also recalls problems with a pneumococcal scheme gene, while expressing uncertainty about the details.
The broader issue is whether bacterial populations form discrete clones. Ed revisits the debate between strict clonality and panmixia, where alleles are extensively shuffled. Neisseria meningitidis illustrated an intermediate, epidemic structure: recognisable clonal complexes within a recombining population. Clones therefore do not necessarily imply little recombination. He contrasts relatively well-behaved S. aureus with more complicated Klebsiella and Vibrio populations, and Burkholderia with substantial allele shuffling despite limited sequence variation.
What MLST leaves behind
For Ed, MLST’s enduring contribution is the nomenclature attached to important lineages. E. coli ST131 is an example: researchers can use the name and understand which lineage is being discussed. Genomic analyses have refined these groupings without overturning many of the central conclusions drawn from seven-gene data.
The method is not simply historical. The discussion notes that some laboratories still lack access to whole-genome platforms, and that some people still submit Sanger traces to PubMLST. Others sequence whole genomes but report only MLST or serological results. Although this leaves much genomic information unused, one of the hosts acknowledges the efficiency of applying the same sequencing pipeline to different organisms.
Highlights
- 00:03:08 — Why the Neisseria MLST scheme settled on seven genes, following MLEE
- 00:06:34 — What MLST measures and why internet-accessible sequence databases mattered
- 00:09:05 — Allelic profiles, sequence types and why nucleotide differences are not event counts
- 00:10:41 — Natacha’s experience using MLST without access to whole-genome sequencing
- 00:12:42 — PFGE’s genome-wide resolution and the practical advantages of sequence data
- 00:19:47 — The origins of eBURST and its model of central founders with single-locus variants
- 00:23:43 — Minimum spanning trees and choosing among alternative connections
- 00:25:52 — Whether early MLST schemes selected suitable genes for every species
- 00:28:57 — The historical debate over clonality, recombination and epidemic population structure
- 00:36:25 — Sequence-type nomenclature as MLST’s lasting contribution
- 00:37:48 — Continued seven-gene typing, Sanger submissions and access to genomic platforms
In their own words
if you look at really, really closely related stuff, everything’s a lot easier.
— Ed Feil, 00:19:47
Who is talking
- Ed Feil (guest, University of Bath)
- Natacha Couto (guest, Center of Genomic Pathogen Surveillance at the University of Oxford)
- Lee Katz (host)
- Andrew Page (host)
- Nabil-Fareed Alikhan (host)
Tools and resources mentioned
Multi-locus sequence typing (MLST), Multi-locus enzyme electrophoresis (MLEE), Pulsed-field gel electrophoresis (PFGE), eBURST, UPGMA, Minimum spanning trees, BioNumerics, PubMLST, MLVA, rep-PCR.
Questions this episode answers
Why do many MLST schemes use seven genes?
Ed explains that six housekeeping genes nearly reproduced the established Neisseria MLEE groupings, but a seventh was needed to distinguish one remaining group. Seven then proved a useful compromise between expense and resolution for other species.
Can an MLST sequence type be assigned with a missing allele?
One of the hosts stresses that a valid sequence type requires an allele sequence for every locus in the scheme. An incomplete profile cannot provide a valid ST.
Why was eBURST developed for MLST data?
Ed wanted an alternative to difficult-to-read dendrograms that represented clonal expansion more directly. eBURST displayed a central inferred founder and related profiles differing at one locus.
Is PFGE more discriminatory than MLST?
Ed says the answer depends on the species: PFGE could detect accessory-genome changes that MLST missed. Natacha used it for additional resolution when investigating possible transmission between animals and humans.
Why is MLST still useful when whole genomes are available?
Sequence-type names such as E. coli ST131 remain useful for communicating about recognised lineages. The episode also notes that some laboratories still lack access to whole-genome sequencing and continue using seven-gene typing.
This page was written from a machine transcript of the episode and checked against it. Listen to the episode for the whole conversation.
Show notes
Ed Feil is a professor of bacterial evolution at the University of Bath, and Natacha Couto, a data scientist at the Center of Genomic Pathogen Surveillance at the University of Oxford. We delve into the concept of multi-locus sequence typing (MLST) in bacterial population genetics.
They highlight how the MLST method allows for defining strains based on partial sequences that range up to 500 base pairs. The method measures differences between loci for each strain, offering an allele number while assigning similar numbers to identical sequences. The cumulative sequence number represents the unique identification, which is subsequently referred to as the sequence type (SST).
MLST has revolutionized the field by facilitating digital storage and comparison of epidemiological databases, proving particularly useful in investigating transmission events and dissemination of certain strains. Although there are other methods such as Pulse Field Gen Electrophoresis (PFGE) that offer higher resolution when looking for similarities between different strains, MLST remains a versatile and widely used method.
They also talk about the shortcomings of MLST and the need for continued improvements in population genetics research. They mention the development of the Eburst program, which uses a circular model, rather than the traditional dendrogram tree structure, to better visualize MLST data and understand the clonal expansion of populations. They also discuss how the original MLST schemes may not have included the best genes for all bacterial species as the genes were chosen before genome sequencing became widely available.
Ed and Natacha further elaborate on the concept of clonality among bacterial species. Ed suggests that bacterial population structures have no consistent pattern, with some organisms being well-behaved, while others have a lot of allele shuffling. However, clones have existed since day one, and their presence is still seen today. Natacha adds that although MLST has flaws, it leaves behind the nomenclature for the lineages or clones, which is a lasting legacy.
Nabil-Fareed notes that while most reference labs have moved on to genomics, some people still use MLST. He adds that the pipeline is the same for any organism, and the process is efficient in the end.
The discussion concludes with the hosts thanking the guests and promising more exciting topics in their next episode. Overall, the hosts highlight the significance of understanding the limitations of MLST and the scope for further research in bacterial population genetics.