Identification of Escherichia coli O157:H7 Genomic Regions Conserved in Strains with a Genotype Associated with Human Infection

Marina Steele, Kim Ziebell, Yongxiang Zhang, Andrew Benson, Paulina Konczy, Roger Johnson, and Victor Gannon

Applied and Environmental Microbiology 73(1), Jan. 2007, 22-31.


I'll presume we all know that E. coli O157:H7 is code for the most common seriously pathogenic form of E. coli currently present in the food supply. The numbers and letters represent the surface 'antigens' that are used to do a preliminary identification of strains of bacteria - the antigens are related to the structures on a bacterial cell surface that are accessible, presumably, to the immune system - but also accessible to assays for testing cell type. Because they are accessible to the immune system, antibodies (particularly monoclonals) can be generated to them and then used in immunoassays, which are very rapid and sensitive. The now classic 'ELISA' (enzyme-linked immunosorbent assay) allows single cells to be identified by their antigens almost immediately. The O antigen is part of the "LPS" of the gram negative cell wall, it is made of sugars. The K antigen is part of the 'capsule' layer - a protective layer of polymerized sugars more loosely connected with the cell. H antigens are derived from the flagella, which are also important to the cell in colonizing a host and are exposed on the cell surface.

Anyway, these antigens can also be used as markers for strongly selected genetic markers - certain antigens are better under certain conditions; particularly, some are more useful for evading predation (diversifying selection acts in this case) and some are more useful for evading the immune system.
Whatever the case, there are hundreds of antigens (~200 O, ~100 K, ~55 H). This allows for tens of thousands of combinations, helping sort the E. coli into clonal monophyletic populations. Of course, it is not necessary that this be true for any given combination, that it is clonal or monophyletic.

Everyone cares about O157:H7 because it keeps showing up in the ill and recently deceased - and in their spinach, tacos, etc. It also shows up in the rectum of ruminants (cattle, sheep) where it doesn't seem to cause the animals any trouble. Certainly, these E. coli have learned some tricks, and modern agriculture encourages their spread. One argument goes that animals being fed grain, not hay, has changed the stomach conditions to permit these E. coli to spread more readily (Diez-Gonzalez et al, 1998; Russell et al, 2000; Russell et al, 2000). The idea is that the E. coli are becoming acid tolerant. Stomach acid is one of the main barriers for humans to prevent infection by bacteria; it protects the established symbionts against temporary invaders. Without going into too much discussion of this point, there is reason to doubt the conclusion that the E. coli are really being selected for acid resistance by cattle diets (Carolyn J. Hovde, Paula R. Austin, Karen A. Cloud, Christopher J. Williams, and Carl W. Hunt, Applied and Environmental Microbiology, 65(7), July 1999, 3233-3235; Grauke LJ, Wynia SA,Sheng HQ, Yoon JW, Williams CJ, Hunt CW, Hovde CJ. Vet Microbiol. 2003 Sep 1;95(3):211-25; Science, April 1999, Vol. 284. no. 5411, p. 49; reviewed, "E. coli O157:H7 in hay- or grain-fed cattle," Dale Hancock and Tom Besser, October 12, 2006, citation not obvious). The evidence for this appears to be more 'suggestive' than 'conclusive.' Anyway, simply lowering the infectious dose may make it easier to become ill, but there must be other things going on to cause O157:H7 to be a problem.

One may be that these E. coli have learned tricks to get into plant roots. There are lots of studies about this issue as well:
JV. Gagliardi and JS. Karns. Environmental Microbiology 2002, Volume 4(2) 89.
Warriner K, Ibrahim F, Dickinson M, Wright C, Waites WM. J Food Prot. 2003 Oct;66(10):1790-7.
The second is interesting in that it presages the spinach problem by showing that E. coli can get internal to spinach cells, so the bacteria could never be washed off. Thus, eating raw spinach fertilized 'organically' with pathogen contaminated manure might create a hazard that cannot be dealt with by surface sterilization.

Anyway, back to the topic of the paper at hand: the relationships among these 0157:H7 bacteria that are causing all this mess.
There are apparently two major families of these bacteria. While even the dominant human pathogens don't seem to cause disease in animals, it seems from most experiments I've seen, that they also don't colonize the animals particularly stably. This doesn't really surprise me - stable intestinal colonization of a conventional (previously naturally colonized by bacteria) host by a laboratory isolate is pretty difficult, in the literature and in my experience. This is true even with large doses of the inocula.

As an aside, infectious doses vary greatly, and it may depend on everything short of the phase of the moon - and even that, in women. One of the great stories of early microbiology was when Koch was trying to demonstrate the germ theory of disease, and he had isolated the cholera causative agent (Vibrio cholera). He had a huge flask of this nasty bug, and another science professor, doubting Koch, postulates and all, decided to demonstrate his scorn. He drank the entire flask of the supposed cause of misery and massive suffering, still one of the most serious diseases - and he didn't get sick. I guess he had taken his vitamins.

Anyway, there are the human-derived Lineage I strains and the bovine Lineage II strains of E. coli O157:H7. they both appear around the world, but one tends to be in the cattle, the other, in disease. They appear to have diverged prior to global dispersal. So, perhaps, all those O157:H7 in cattle that people are all stressed about aren't a big deal... The authors in this paper decided to figure out what genes were different among the lineages. They found a bunch of regions that were different, and then characterized them to look for virulence factors which may make Lineage I a problem and Lineage II a bunch of bullcrap, literally.

They used 30 strains for the secondary screen - 10 Lineage I and 20 Lineage II, to confirm the differences found in a smaller primary set of 4 strains. Then they did a tertiary screen with 119 additional strains.

They found a bunch of regions that were relevant. The big finding is that most of these were associated with potential viruses, plasmids, and transposons - mobile genetic elements that can thereby change their background and associations. This begins to explain the apparent recent origin of the pathogens. Some of the genes look like obvious problems - colonization elements, hemolysin genes, outer membrane molecules. Others are not obvious.

One thing that this says is that these factors might reassemble in different combinations in other backgrounds. O157:H7 might have just assembled all the cool bad genes, a sort of Ocean's 11 of E. coli, all at once.

Another interesting thing is that it looks like 'lineage II' is just lineage I that has lost elements - perhaps there is selection against virulence elements in the cattle. Perhaps there is general pressure to be less virulent, not more. this would be very good for us and our food supply.

This is a very cool article on molecular ecology and evolution in the context of pathogenesis and symbiosis. It makes for some cool findings and begins to suggest some interesting approaches for prevention and therapy. For instance, curing bacteria of phage would likely be enough to eliminate many virulence factors without killing the cells. Another interesting point - if we can keep bacteria from being infected with phage, they might not emerge to virulence, if that is a major source of virulence factors which are otherwise being purified by selection.

Identification of Escherichia coli O157:H7 Genomic Regions Conserved in Strains with a Genotype Associated with Human Infection

Marina Steele, Kim Ziebell, Yongxiang Zhang, Andrew Benson, Paulina Konczy, Roger Johnson, and Victor Gannon

Applied and Environmental Microbiology 73(1), Jan. 2007, 22-31.


I'll presume we all know that E. coli O157:H7 is code for the most common seriously pathogenic form of E. coli currently present in the food supply. The numbers and letters represent the surface 'antigens' that are used to do a preliminary identification of strains of bacteria - the antigens are related to the structures on a bacterial cell surface that are accessible, presumably, to the immune system - but also accessible to assays for testing cell type. Because they are accessible to the immune system, antibodies (particularly monoclonals) can be generated to them and then used in immunoassays, which are very rapid and sensitive. The now classic 'ELISA' (enzyme-linked immunosorbent assay) allows single cells to be identified by their antigens almost immediately. The O antigen is part of the "LPS" of the gram negative cell wall, it is made of sugars. The K antigen is part of the 'capsule' layer - a protective layer of polymerized sugars more loosely connected with the cell. H antigens are derived from the flagella, which are also important to the cell in colonizing a host and are exposed on the cell surface.

Anyway, these antigens can also be used as markers for strongly selected genetic markers - certain antigens are better under certain conditions; particularly, some are more useful for evading predation (diversifying selection acts in this case) and some are more useful for evading the immune system.
Whatever the case, there are hundreds of antigens (~200 O, ~100 K, ~55 H). This allows for tens of thousands of combinations, helping sort the E. coli into clonal monophyletic populations. Of course, it is not necessary that this be true for any given combination, that it is clonal or monophyletic.

Everyone cares about O157:H7 because it keeps showing up in the ill and recently deceased - and in their spinach, tacos, etc. It also shows up in the rectum of ruminants (cattle, sheep) where it doesn't seem to cause the animals any trouble. Certainly, these E. coli have learned some tricks, and modern agriculture encourages their spread. One argument goes that animals being fed grain, not hay, has changed the stomach conditions to permit these E. coli to spread more readily (Diez-Gonzalez et al, 1998; Russell et al, 2000; Russell et al, 2000). The idea is that the E. coli are becoming acid tolerant. Stomach acid is one of the main barriers for humans to prevent infection by bacteria; it protects the established symbionts against temporary invaders. Without going into too much discussion of this point, there is reason to doubt the conclusion that the E. coli are really being selected for acid resistance by cattle diets (Carolyn J. Hovde, Paula R. Austin, Karen A. Cloud, Christopher J. Williams, and Carl W. Hunt, Applied and Environmental Microbiology, 65(7), July 1999, 3233-3235; Grauke LJ, Wynia SA,Sheng HQ, Yoon JW, Williams CJ, Hunt CW, Hovde CJ. Vet Microbiol. 2003 Sep 1;95(3):211-25; Science, April 1999, Vol. 284. no. 5411, p. 49; reviewed, "E. coli O157:H7 in hay- or grain-fed cattle," Dale Hancock and Tom Besser, October 12, 2006, citation not obvious). The evidence for this appears to be more 'suggestive' than 'conclusive.' Anyway, simply lowering the infectious dose may make it easier to become ill, but there must be other things going on to cause O157:H7 to be a problem.

One may be that these E. coli have learned tricks to get into plant roots. There are lots of studies about this issue as well:
JV. Gagliardi and JS. Karns. Environmental Microbiology 2002, Volume 4(2) 89.
Warriner K, Ibrahim F, Dickinson M, Wright C, Waites WM. J Food Prot. 2003 Oct;66(10):1790-7.
The second is interesting in that it presages the spinach problem by showing that E. coli can get internal to spinach cells, so the bacteria could never be washed off. Thus, eating raw spinach fertilized 'organically' with pathogen contaminated manure might create a hazard that cannot be dealt with by surface sterilization.

Anyway, back to the topic of the paper at hand: the relationships among these 0157:H7 bacteria that are causing all this mess.
There are apparently two major families of these bacteria. While even the dominant human pathogens don't seem to cause disease in animals, it seems from most experiments I've seen, that they also don't colonize the animals particularly stably. This doesn't really surprise me - stable intestinal colonization of a conventional (previously naturally colonized by bacteria) host by a laboratory isolate is pretty difficult, in the literature and in my experience. This is true even with large doses of the inocula.

As an aside, infectious doses vary greatly, and it may depend on everything short of the phase of the moon - and even that, in women. One of the great stories of early microbiology was when Koch was trying to demonstrate the germ theory of disease, and he had isolated the cholera causative agent (Vibrio cholera). He had a huge flask of this nasty bug, and another science professor, doubting Koch, postulates and all, decided to demonstrate his scorn. He drank the entire flask of the supposed cause of misery and massive suffering, still one of the most serious diseases - and he didn't get sick. I guess he had taken his vitamins.

Anyway, there are the human-derived Lineage I strains and the bovine Lineage II strains of E. coli O157:H7. they both appear around the world, but one tends to be in the cattle, the other, in disease. They appear to have diverged prior to global dispersal. So, perhaps, all those O157:H7 in cattle that people are all stressed about aren't a big deal... The authors in this paper decided to figure out what genes were different among the lineages. They found a bunch of regions that were different, and then characterized them to look for virulence factors which may make Lineage I a problem and Lineage II a bunch of bullcrap, literally.

They used 30 strains for the secondary screen - 10 Lineage I and 20 Lineage II, to confirm the differences found in a smaller primary set of 4 strains. Then they did a tertiary screen with 119 additional strains.

They found a bunch of regions that were relevant. The big finding is that most of these were associated with potential viruses, plasmids, and transposons - mobile genetic elements that can thereby change their background and associations. This begins to explain the apparent recent origin of the pathogens. Some of the genes look like obvious problems - colonization elements, hemolysin genes, outer membrane molecules. Others are not obvious.

One thing that this says is that these factors might reassemble in different combinations in other backgrounds. O157:H7 might have just assembled all the cool bad genes, a sort of Ocean's 11 of E. coli, all at once.

Another interesting thing is that it looks like 'lineage II' is just lineage I that has lost elements - perhaps there is selection against virulence elements in the cattle. Perhaps there is general pressure to be less virulent, not more. this would be very good for us and our food supply.

This is a very cool article on molecular ecology and evolution in the context of pathogenesis and symbiosis. It makes for some cool findings and begins to suggest some interesting approaches for prevention and therapy. For instance, curing bacteria of phage would likely be enough to eliminate many virulence factors without killing the cells. Another interesting point - if we can keep bacteria from being infected with phage, they might not emerge to virulence, if that is a major source of virulence factors which are otherwise being purified by selection.

Use of 16S rRNA and rpoB Genes as Molecular Markers for Microbial Ecology Studies

Rebecca J. Case, Yan Boucher, Ingela Dahllöf, Carola Holmström, W. Ford Doolittle, and Staffan Kjelleberg

Applied and Environmental Microbiology, 73(1), Jan. 2007, 278-288

Here we have a follow-up study to one discussed earlier in this blog, about the troubles of using 16s rRNA for phylogenetic studies.

16s rRNA is a very slowly evolving molecule. There are a number of reasons for that. One is that the 16s secondary structure means that most mutations require a second mutation or else it interrupts a pairing pattern and ruins the structure. Second, the 16s is part of the huge ribosome complex that is really key to most of life's processes (central dogma stuff). You simply can't live if your ribosomes are mucked up, even to the point that the run slowly. The 16s has to interact with all sorts of proteins and other rRNAs. Most of it is critical, one way or another. On top of that, most bacteria have multiple rRNA genes. They need to be able to control their investment in ribosomes and also produce lots and lots of them quickly. Thus the copies are regulated so that some are almost always on, while others turn on only when growth is very rapid. This allows for great flexibility in growth rate. Having multiple copies allows for 'correction' by copying a good copy onto a bad copy. This slows evolution; it may relax purifying selection a little in the copies that are more rarely used, but it also ups the ante on muller's ratchet. Multiple copies, variable per cell, also means that a single copy of the gene doesn't mean a single cell - or even a linear relationship. This is a very serious problem for doing relative counts among different uncultured bacteria.

Anyway, between having multiple copies in the same cell, which may vary at the few points that variability is allowed, and having little variability to start, this creates a mess for phylogeny. Still, 16s is the gold standard, in part because of the size of the database. Also, it is ubiquitous (good) and it has good primers (good) and it is generally accessible to polymerase (good; other genes may have secondary structure in the chromosome that makes copying the gene difficult); further, multiple copies in a cell means there are more targets, more to work with per unit DNA (super for FISH - but differing accessibilities of differing copies complicates this picture as well). It is also an RNA, so there is no 'codon bias' or 'wobble' complications. Still, this way of thinking really disguises the complexity of the basis of mutation rate variation across a functional RNA molecule, which is at least as complicated as that for proteins.

Various genes have been suggested to augment or replace 16s rRNA for various taxa. RecA for Vibrio is one example. MLSA is another option (Santos, S. R., and H. Ochman. 2004. Environ. Microbiol. 6:754-759; Thompson, F. L., D. Gevers, C. C. Thompson, P. Dawyndt, S. Naser, B. Hoste, C. B. Munn, and J. Swings. 2005. Appl. Environ. Microbiol. 71:5107-5115.), and then there is the nice work by Konstantinos T. Konstantinidis, Alban Ramette, and James M. Tiedje (Applied and Environmental Microbiology, November 2006, p. 7286-7293, Vol. 72, No. 11).

In this paper, they suggest RpoB as a single universal gene. Of course, it carries all the baggage of using any protein coding gene. However, it is easier than trying to use a large set of genes - which, though it gets cheaper all the time - is still very expensive, very difficult, and impossible to do in a 'metagenomic' way by just pulling genes out of mixed template without 'coordination' - getting clones or other identifiable genomic units that can be linked together.

From their evidence, it seems that the choice of RpoB is as good as any, but there is no great evidence that it is uniquely better than either another gene, or some collection of genes that changes based on the overarching taxonomic framework that uses different markers for subgroups and something like 16s to generate the coarsest classifications.

At the end of the paper, the authors discuss the use of functional genes - using a gene as a marker not for a taxa, but for a capability (like phenol degradation, etc. This is naturally hugely complex, especially as many certain capabilities can be achieved multiple ways, using unrelated genes. This notion is not advanced, it seems to me, by using RpoB, unless I'm missing something. My bet is that this methods paper is really a pre-paper to carry the methods for something more interesting coming down the pike. Still, it represents an interesting body of data describing how replacing 16s is complex, and providing some more data on how this may or may not be accomplished.

Shiga Toxin and Shiga Toxin-Encoding Phage Do Not Facilitate Escherichia coli O157:H7 Colonization in Sheep

Nancy A. Cornick, Amy F. Helgerson, and Vijay Sharma
Applied and Environmental Microbiology, 73(1), Jan. 2007 344-346

In this paper, the big question of the prevalence of shiga toxin bearing Escherichia coli in the ruminant population is addressed. Shiga Toxin genes are generally carried by a phage (bacterial virus) that integrates into the E. coli genome. These genes are a major factor in creating a serious human pathogen from an otherwise benign intestinal commensal. In other words, bad bacteria from good bacteria.

Basically, these shiga toxin carrying bacteria don't cause any noticeable problem for ruminants - fore-gut fermenters - as opposed to what they do to humans. However, they are carried at relatively high rates by ruminants. This brings them into the food supply both as contaminants of meat, and then, as manure, as contaminants of organic produce.

So, why are they so prevalent? What about ruminants and shiga toxin creates such a high occurrence, by which many cattle and sheep carry these human pathogens?

In this paper, there is an attempt to see whether the shiga toxin creates higher rates/longer periods of shedding by sheep. It is a proper and elegant study, in that the parent strain (a human pathogen, taken from a clinical case) carries the phage and is then subjected to mutagenesis, to remove the toxin alone, or also the phage as a whole. They inoculated different sheep with different strains, and even tried co-inoculations.

They did a power study, and determined that they had an 80% expectation of finding a real effect (if there was one) using 8 sheep per group. So, that's what they did. This feels like a serious footnote - there is a 20% chance that their conclusions are wrong. However, this sort of analysis really belongs in many more studies than have it.

First, we should highlight that _all the inoculated strains_ disappeared over time from the animals. They were using antibiotic resistance as markers and running the experiment for several months under non-sterile conditions (but BL2 to prevent escape of the pathogen). The animals already had living flora. So, we are really looking at differential loss among losers, not loss or success among something that would be a stable floral component. The experiment only went a couple months; a stable strain should really last longer than that.

They also only sampled shedding - not whether the animals were carrying the bacteria internally - and they acknowledge this weakness in the text. They didn't look at colonization dosage, either, or movement of the toxin from strain to strain, or at animals who were coinfected with BLV, testing a variety of hypotheses about the effects of the shiga toxin. They also didn't look at survival in feces or anything about that part of the bacterial ecology cycle.

However, the one hypothesis they sought to test, they found no particular effect. There was lots of variability among individuals and no distinguishable bias towards or away from the shiga carrying strains. Certainly, it would have been interesting to see an effect.

This doesn't necessarily mean that the toxin is 'selfish' and not helping the bacteria; however, it may mean that the effect is really found elsewhere in the life cycle, or requires other genes to be coordinated with it. It might be that there is a population of bacteria that have some genes where shiga helps them colonize ruminants, and that the phage is shed by these bacteria and infects O157:H7 type strains, producing a human pathogen, while the human pathogen itself isn't favored in the animals. Then again, it might be something completely different. We need more good little experiments - hopefully, with a slightly higher power than 80%, but we'll take what we can get - to find out how this all works.