If ever you have suffered from acid reflux, then you may have been prescribed a "proton pump inhibitor". These act, as you may have been able to guess, on proton pumps within the gut. so what do proton pumps do ?
Proton pumps keep your stomach acidic
In order to keep your stomach acidic, proton pumps act to pump protons (also known as hydrogen ions) into your gut. But in cases where your gut produces too much acid, doctors can prescribe proton pump inhibitors such as Omprezole. They stop the proton pumps, and thus prevent the build up of stomach acid. This de-acidification of the stomach can leave someone vulnerable to gastric infections. A systematic review has shown that proton pump inhibitor treatment is associated with infections by bacteria such as Salmonella, Campylobacter, and Clostridium Difficile *.
We are focussing on C. difficile today. There are many types of C. difficile, but the ones we want to pay attention to are the toxigenic strains, which can produce toxin A and Toxin B. These are encoded by the tcdA and tcdB genes respectively These are transcribed by a special sigma factor, known as tcdR. The tcdC gene is a regulator, preventing these genes from being expressed until stationary phase. In some "hypervirulent" genotypes of C.difficile, such as 027**, 078*** have deletions in the tcdC gene, which some researchers believe renders it non-functional in these genotypes, thus explaining their "hypervirulence".
But there is a vast soup of other factors that perhaps could affect virulence gene expression, from the pH of an environment, to nutrient availability. Could proton pump inhibitors also affect virulence genes of C. difficile ?
What did they do ?
The researchers took samples of C.difficile from three patients infected with different types of C.difficile, type 001, 027 and 078. They then grew samples of these bacteria in brain heart infusion broth with different concentrations of Sodium Hydroxide and Hydrochloric acid, to ensure they have different pH's. When the bacteria enter stationary phase, the researchers portioned off separate samples which they treated with either 0 or 100 or 200 microlitres of Omeprazole for either four or twelve hours. At the end points, the cells had all of their RNA extracted.
From this soup of RNA, the researchers searched for RNA sequences related to the virulence genes. This brings us to Figure 1.
I am presenting this graph at the original resolution it was presented to me. This essentially shows the effects of omeprazole and pH on C. difficile transcription of Toxin A. We can see that the biggest changes in Toxin A expression occur at pH 9, which if it occurred in your body probably means you are the pillsbury dough boy and have baking soda for blood. But I'm glad they showed that data, because it makes the data they actually want to present impossible to see properly. No, wait, I'm not.
Toxin B behaves somewhat differently. This time, we see that there are some differences between the strains. My favourite part of this graph is how all the highest results have been clustered together for the sole purpose of making it hard to discern where 001 ends and 027 begins.
C. difficile strains 027 and 078 also produce binary toxin, another important virulence factor. So we can see peaks occurring on those two graphs, the highest appearing at pH5 in 027. My favourite part of this graph is how the graph lines for most of the data for 027 don't sync up with the rest of the graph, making it look like it was cut and pasted in using MSpaint. Truly, we are witnessing the apotheosis of data presentation here.
They next looked at the virulence regulators, which you can find in the supplementary results. I'm literally just going to show you the results graphs, and comment on them afterwards.
Promoters of virulence
Repressors of virulence
You'll notice that the highest transcription in all of these groups is always in the 027 group at pH5, for both the activating genes, and the repressing genes. Why would this happen ? How could there be an increase in the genes making more toxin, and the genes making less toxin ?
I think there is a very simple answer to this question, and one that put the whole paper in perspective for me. But I have a number of things I want to talk about with this paper.
I'll begin with a caveat, but not my own. The authors themselves make very clear in this paper:
There are teeth marks on my keyboard because of that statement. To explain why, allow me to translate the above statement for you.As this study dealt with technical replicates, as opposed to biological replicates, measurements of fold-change for gene expression were utilized, and statistical comparisons between technical replicates was not performed as this was not felt to be mathematically or methodologically appropriate.
"we are operating with an n=1 for all of our experiments. We didn't do statistics, because that would be laughable."But that isn't even what irks me the most about this statement. No, It's not the sheer brazenness of it, the insult to the intelligence of every reader of this paper. It's the implication that if you put "Technical replicate" in front of anything, you get a free pass for not doing any actual replicates.
I'm not even getting near this paper's biggest problem.
Real Time PCR- How does it work ?
I mentioned that at the end of every growth experiment, the researchers measured the RNA transcripts of their target genes using Real Time PCR. But Real-Time PCR only works when you have a gene to compare against. Essentially, you need a control gene, one which doesn't change its expression in response to any of the experimental treatments you are testing. If say, you choose a control gene that isn't transcribed at low pH's, then even if the gene you're measuring doesn't change, it will look like its going through a spike, or if your gene is changing, then those changes will be masked by the actions of your control gene.
Usually, researchers use housekeeping genes, such as 16s RNA , which are supposed to be transcribed at the same rate for most environments.
Except, that this isn't always the case. It is known that when exposing bacteria to severe stress, the amount of 16s RNA can vary wildly. In a situation in which the bacteria are being exposed to acid over a long term time course would certainly qualify as stressful. The choice of 16s RNA needs to be validated for this experimental protocol.
You see, if you look at the last set of data knowing this, you could look at the last four figures in this study, and draw the conclusion that 027 loses it's 16sRNA more quickly compared to the other strains tested in the presence of acid. And let us not forget that some strains sporulate more quickly than others, and so a proportion of the bacteria obtained at the end of an experiment may have formed endospores, and could skew the results. This is a massive problem. Since all of the results are completely skewed, it is impossible to draw any conclusions from this paper.
I mean, I guess if there were meaningful replicates we could deduce a pattern in the data, and perhaps try to account for this skew. But wait, there aren't.
You may have wondered why I was so late in posting this write up about the paper. The truth is that I really agonised over whether I should publish this blog post, or even take part in the next #microtwjc. Someone made the effort of making this article open access just for today's #microtwjc, and it genuinely upsets me that I can't say something nice about this paper in return. But it is just SO BAD.
The graphs were an abomination unto nature. The writing style appeared to have been designed to hit all of my berserk buttons. The methods were so completely flawed as to render the paper meaningless.
Stewart D.B. & Hegarty J.P. (2013). Correlation between virulence gene expression and proton pump inhibitors and ambient pH in Clostridium difficile: results of an in vitro study, Journal of Medical Microbiology, 62 (Pt_10) 1517-1523. DOI: 10.1099/jmm.0.059709-0
EDIT: SGM have published a new version of the paper, which corrects the graphs at least to make them readable. But I currently award it no points, because the data on Fig 1 is exactly the same as Fig 2, because they decided to erase all of the tcdA data. I've alerted them to this, and hopefully the publishers will sort it out soon.
*Systematic review can be found here