As part of our educational programme, we have been looking at EAE experiments.
Education: Understanding EAE. The Reverse
Experimental Autoimmune Encephalomyelitis (EAE) often gets a bad rap. Often it is not the fault of the model but the way it is used and importantly interpreted. However, some of it, a small amount, puts the c in rap and we are not talking Dr. Dre here.
It is useful for anyone reading EAE papers that they understand how to interpret the data. This is especially important for people working in the media. As over interpretation, leads to false hope.
Disease course in Biozzi ABH mice Mouse EDSS
Part of the difficulty of interpreting what is going on, is the fact that not enough information is reported.
We are not told what the score of the animals is, what is the day of onset, what is the number in a group or importantly the number of animals that got disease.
Is it a good experiment? or something that should be repeated? because it lacks enough Quality Control!
A line graph can mean a lot or a little.
Without extra information they can be mean different things
If you have a mean score of one, you have to ask how did we get that?
1 = limp tail, 2= altered gait, 3 = partial paralysis, 4 = paralysis, 5 =dead (See above)
1 + 1 + 1 + 1 + 1 + 1 + 1 + 1 Mean =1
(All show the same effect)
0 + 0 + 0 + 0 + 0 + 0 + 4 + 4 Mean =1
(Is this a duff experiment where only 2 animals got disease)
3 + 3 + 3 + 3 + 3 + 3 + 3 + 3 Mean = 3
5 + 5 + 5 + 5 + 4 + 0 + 0 + 0 Mean = 3
(4 dead + 3 no disease)
I think the low control score of around one usually means it is a duff experiment! However this often slips through the refereeing process.
To interpret line graphs we need to look for and report deviation in groups.
What does this mean? The green line is lower than the pink line so it looks like a good treatment effect. With no indication of the deviation of the result and how many animals are contributing to the score it is difficult to decipher.
With a “small deviation” meaning most animals are showing the same level of disease, the drug may be having a significant effect. It could be interesting and the next new treatment for MS.
With a “large deviation” the scores of animals are very different and overlaps are common and so this may mean the study is uninteresting, the drug does not work or may be the control group has not worked well enough to make any conclusions and the study needs to be done again.
It is poor quality reporting and poor refereeing, if this information is missing in papers.
Lack of showing deviations usually means the data is too messy….
and does not give you as much confidence that the data is biologically meaningful or that it will be repeatable…
Likewise we are often not told when animals get disease onset.
The type of line graph (above) has been called by some as “progressive EAE” (I don’t buy this) because the line graph shows that the neurological score gets slowly worse. In contrast the disease profile (Below) in EAE in C57BL/6 is often called “chronic EAE” because they do not recover. This is because of nerve loss leading to irreversible damage.
However, in the above cases it could be reproducible disease verses inconsistent disease and the two graphs were based on the same scores, one where the onset was within a few (two) days, the other over a protracted time over 12 days
The scores of 10 mice are shown as above
This should give you a view that without knowing the day of onset, the severity of affected animals, etc. it is difficult to really know what is going on.This information therefore needs to be reported.
Many, many times we see a drug treatment effect which inhibits disease the control mean score of say 3 and the drug mean score is 1. It could be a wildtype (normal) verses a transgenic mouse, which adds the inhibitory molecule rather than a drug study.
So now we do the experiment in reverse and we put in an inhibitor of the drug target or use a gene knockout and we have the placebo
with a score of 1 and the knockout group score 3. The blocker makes things worse.
Great….Lets get our paper in Nature, Immunity Science etc.
On face value the data all looks great, and we accept the science.
However, the only thing that has changed is how good the disease in control group is!!!. Therefore, is the data correct? or is it a chance effect?.
If it goes the right way, are you happy and do you bother to repeat the experiment?
Does this happen? Sure it does! Next time you read/referee a paper, look out for the variation in the score of the control group. Differences happen quite often, especially in EAE in C57BL/6 mice.
What we want is Quality Control so the control group gives consistent disease. Quality control within the system tells you if the system is working and if it does not, you should think that it needs to be repeated.
We want data to look more like this
The control groups same intensity = More confidence in Result
When you read many EAE papers there are often issues in reporting quality and sometimes issues with Quality Control of the results.
This will be part of the failure to translate and small influences on EAE may have little biological relevance to treatment of MS, but is enough to convince referees to accept the publication.
As animals are inbred i.e. genetically identical if means the experiment is being done in the same individual so you would expect little deviation in results. Is this individual representative of what is happening?
Unfortunately, work is increasingly being concentrated in C57BL/6 mice. This is because it is the background strain used in transgenic and gene knockout mice has become dominant. It is however one of the most genetically-EAE resistant strains….It is also a strain that will preferentially drink alcohol compared to water..must be British:-)
I suspect that many ideas based on results in this strain, will fail to translate to other strains let alone humans as the disease quality in this genetic background is inconsistent…….Call me a cynic!
If you have access to the data you can work it all out,
However surely it would be better for the referees to have access to this so they can rectify mistakes before they are published, rather than have the Science Police spot the errors.
The Journals would then need to store the information