How Not to Isolate E. coli

If you’re interested in studying what ‘real’ bacteria do–that is, not the microbiological equivalent of anemic inbred lab rats–you have to isolate them from natural samples. Yogi Berra-esque statements notwithstanding, how you isolate bacteria is critical. Methodological details do matter. Which brings me to a recent paper that makes me very cranky (though not Mad) about antibiotic-resistant E. coli isolated from racoons (boldface mine):

E. coli isolation. Fecal specimens ranged from 0.5 g to 5 g. A portion of the specimen was suspended in buffered peptone water (Becton Dickinson, Sparks, MD) at a 1:10 ratio and incubated at 37°C overnight. The broth suspension was homogenized using a vortex, and 2 ml was dispensed into an equal amount of double-strength EC (Escherichia coli) broth (Becton Dickinson) and incubated at 37°C overnight. Following incubation, a loopful of cultured EC broth was inoculated onto MacConkey agar (Becton Dickinson) and incubated at 37°C overnight.

To translate this into English, they took some diluted raccoon shit and let the bacteria (mostly E. coli and relatives*) overnight in a nutrient broth. They then took some of that overnight culture and and let those bacteria grow overnight again.

We are very cranky. Why?

First, as a matter of good technique, you should always try to have the isolation procedure not affect the distribution of strains. In other words, suppose my sample has two E. coli strains (genetically distinct E. coli), A and B, and A is much more abundant than B. Now suppose B grows faster than A. At the end of this process, I will have a lot more B than A. If I’m only picking a few colonies, which results in essentially sampling the most common strain (B), I have a completely backwards picture of what is actually in my sample. A related problem is that you’re missing bacterial diversity you would otherwise capture.

But it gets worse. Some bacteria produce what are known as bacteriocins, proteins that kill closely-related bacteria, in order to gain a competitive advantage (in E. coli, they’re called colicins). In a flask, after overnight growth, the colicin-producing strain will all but wipe out the colicin-sensitive strain. This isn’t conjecture: I spent many years of my life doing exactly that (SCIENCE!).

Not done yet though. Many strains of E. coli carry bacteriophage, which are bacterial viruses. These bacteriophage can infect other E. coli strains. This can, in some cases, wipe out the uninfected strain (similar to the colicin case above). However, in other cases, the bacteriophage integrate into the bacterial genome (kinda like a retrovirus, although the mechanism is different). Let’s return to our friends, A and B, but this time B carries a phage that infects strain A. Now we’ve created a strain (‘A-phage’) which doesn’t exist in the sample. Oops.

Which brings us back to the article’s topic: antibiotic resistance. Many antibiotic resistance genes are found on mini-chromosomes, called plasmids, that can jump from strain to strain. Guess what often happens when you put two strains together, one which has a plasmid and one which doesn’t? They move. Again, you’ve created a new strain that doesn’t exist. The saving grace of the study is that there wasn’t any selective isolation: they didn’t then kill off all of the antibiotic-sensitive bacteria, so the odds of picking the ‘new’ strain are low (unless, of course, the plasmid does something advantageous…like produce a colicin).

But I really don’t like this kind of isolation procedure because I don’t trust what you found–if you ever want to do something else with these isolates, like understand what E. coli live in raccoons, you have no idea what was actually in your sample.

It’s one thing to suspend your sample for a brief period in a dilute nutrient solution to allow the bacteria to recover–plating media can be very harsh. And microbiology has always suffered from an isolation problem–no method perfectly captures what’s in the sample. But two overnight competitive bottlenecks is too much. In this case, it probably isn’t a fatal flaw, but it definitely limits the future relevance of this collection.

Cited article: Jardine CM, Janecko N, Allan M, Boerlin P, Chalmers G, Kozak G, McEwen SA, Reid-Smith RJ. 2012. Antimicrobial Resistance in Escherichia coli Isolates from Raccoons (Procyon lotor) in Southern Ontario, Canada. Appl Environ Microbiol. 78(11):3873-9.

This entry was posted in Antibiotics, E. coli, Microbiology, Viruses. Bookmark the permalink.

2 Responses to How Not to Isolate E. coli

  1. ele says:

    But isn’t the same thing looking for a bacteria in a sample than studing the ecosystem of a sample, no?

  2. Lisa Gorski says:

    As someone who researches pathogenic bacteria in foods and in the environment, I agree with your post. I’ve published a couple of papers on the bias that enrichment protocols can infer on the isolation of Listeria and Salmonella leading to the selective enrichment (or not) of a particular strain of genera over others. AEM 2006; 72:776 and PLoS ONE 2012 7(4): e34722.

    The problem in transferring this to real world applications in my research in food safety is that often in real life samples you need to enrich to see anything. Numbers are too low in the sample. I’m currently studying samples with higher contamination to see if enrichment protocols bias what comes up in these naturally contaminated samples.

    However since I do assess antibiotic resistance in some of our isolates now I have more to worry about.

Comments are closed.