[Neurosciences/BBB] Journal Club – Blood–brain barrier and intestinal epithelial barrier alterations in autism spectrum disorders

An interesting paper that was recently published by Fiorentino and colleagues entitled “Blood–brain barrier and intestinal epithelial barrier alterations in autism spectrum disorders”  in Molecular Autism journal. The article is in open-access so you can download the full-article here: (https://molecularautism.biomedcentral.com/articles/10.1186/s13229-016-0110-z).

In this paper, the used postmortem brain tissue and duodenum tract biopsies samples from patients diagnosed with autism spectrum disorders (ASD), schizophrenia (SCZ) and compared to controls. They measured changes in tight junctions and inflammation at mRNA and protein levels using qPCR and Western-blot respectively.
One aspect that I found missing in the sample used is the size of the male population, largely bigger than the female population. It would have been neat to have the same population weight (as many male samples than female samples) but you cannot choose the samples you get. Also, the ethnicity is predominantly Caucasian.
The data is interesting but also a bit confusing in terms of statistical analysis, with some p-hacking and data misleading by showing the standard error mean (SEM) instead of standard deviation (SD). The SEM (SEM=√SD/number of samples) has little meaning when your sample size below 100. it just makes the graph nicer, but it does not help the reader to assess the statistics.
At mRNA levels, the authors found an increase in claudin-3, claudin-5, claudin-12 and tricellulin (TRIC) mRNAs in ASD patients compared to SCZ. Now here what bothers me is their selective statistical analysis, almost cherry-picking it. For instance, claudin-3 is increased in both the SCZ and ASD groups. They observed a statistical difference between ASD and control, but mentioned nothing about the SCZ versus control, despite showing similar mean +/- SEM. Now Claudin-5, claudin-12 and tricellulin (TRIC) were selectively increased in ASD patients compared to control, but again they failed to compare SCZ to control samples (side-note: the authors failed to provide the primer sequences for TRIC. Bad job!). Still it is interesting, tight junction (TJ) complexes seem to be up-regulated in ASD patients compared to SCZ and controls. Does it mean the BBB is compromised? At this point, no. Seems even the BBB maybe better.
In the next figure (Fig.2) there are some interesting data but also some dishonest statistics done. The authors consistently compared ASD to SCZ instead of controls. Thats some scientific dishonesty that should have been caught by reviewers. Shame on you reviewers!
Still there are some interesting trends. The authors observe an increase in matrix metalloproteinases (MMP-2 and MMP-9). These are known proteases at the BBB are they are commonly associated with a loss of the BBB during stroke, these MMPs can cleave the basement membrane supporting the BBB and also the TJ complexes. But not much for the other markers used, except maybe a down-regulation in SCZ patients compared to controls. The same sort of p-hacking occurred for looking at inflammatory cytokines (IL-1beta, IL-6 and IL-8). There is not much differences between ASD versus control, but by using the Texas sharpshooter fallacy (by using the SCZ as a “control”) you bring statistical differences where there is not. This is the big problem of this paper: trying to oversell the data using a p-hacking method.
Now expression at mRNA level is not 100% translated in changes at protein levels. in Figure 5 and 6, the authors looked at the changes at protein levels for claudin-5 and claudin-12 The authors used two protein controls: alpha smooth muscle actin (SMA) and beta-actin (ACTIN). ACTIN is a housekeeping gene, it is expressed in all cells at very high level. This is why we use ACTIN as a loading control, to be sure we are observing differences solely based on protein expression not due to differences in amount of total proteins between samples. The use of SMA is smart because it can help determine the amount of medium and large blood-vessels. Capillaries (the vessels carrying on the BBB phenotype) are poor in SMA expression, as they have few smooth muscle cells (SMA is a smooth muscle cell marker). In vessels with higher diameter, you have a intermediate layer (called media) rich in smooth muscle cells. The blots are interesting but I thought the normalization was not great. The authors normalized Claudin-5 and -12 expression against SMA. I would have first show the normalization of these proteins and SMA against actin to have an overall view and show that our samples are not biased (if our samples are random, so should be SMA. However, we are expecting to see a relative similar levels of SMA between the different groups). Interestingly, Claudin-5 protein level is higher in ASD compared to control, whereas claudin-12 is lower.
Figure 7 is in my opinion interesting and confusing in the same time. The reason it is confusing it the experimental design. I have no clues on what the author try to show us. They show a panel of different mRNA expression of different markers for each patients separately. They should have kept the same representation (the average expression of each individuals plotted as dot-plot) as the previous figure. My speculation is that once average, they were facing a important variability in the sample distribution that nothing was likely to come out (likely no observable differences compared to healthy controls). Almost feel like the authors have these piece of extra data with no where to go and used to fill up the paper. It may also explain why the authors failed to show any proteins for these data.
Now, the authors are a bit over-optimistically concluding that there is a leaky BBB in ASD patients. That’s not exactly true for two reasons: claudin-5 levels are higher and they fail to provide adequate data to support this claim. Post-mortem immunostaining against TJ complexes can let measure changes in TJ strands in microvessels. By measuring the average length of these strands, you can speculate a disrupted BBB. In addition, non-invasive MRI (with gadolinium tracer) would help refine this claim and demonstrate evidence of this claim.
The idea of a disrupted BBB in ASD patients is interesting and some of these data are interesting. Yet this study has some issues to convince a BBB expert about the presence of a leaky BBB. Dont let get fooled by some snake oil sellers that will use this paper to support some “leaky brain” or “leaky gut” in ASD patients. This paper show anything except these bogus claims.


2 thoughts on “[Neurosciences/BBB] Journal Club – Blood–brain barrier and intestinal epithelial barrier alterations in autism spectrum disorders

  1. No “dishonest” statistical analysis was done. Both SCZ and ASD samples were always compared to HC. The one-way ANOVA is meant to compare more than 2 groups and the authors performed statistical analysis on 3 groups. SCZ were NEVER used as controls in this paper. HC were the reference controls. SCZ group was meant to be a further control for disorders affecting the brain.
    “side-note: the authors failed to provide the primer sequences for TRIC. Bad job!”…Authors listed primers sequences in a separate table. TRIC as other primers were purchased from Qiagen which do not release their sequences.
    Maybe as more attentive reading of the paper would have helped.

    Liked by 1 person

    • Oooh, interesting to see the first author commenting on my blog post. I will let your comment public. In Figure 3, can you explain why you don’t show the statistics of IL-1b, IL-6 and IL-8? You have a statistical differences versus SCZ (that is nice) but I was more eager to see what was going on versus control (since you have no asterisks, I assume no statistical differences).
      Also there are some big differences in the SD/SEM sizes between groups. Homoscedaticity is important to maintain in ANOVA, this is a cardinal rules (n independent, same theoretical SD, follow normal/Gaussian distribution). For primers, you can rely on QIAGEN but in a time when rigor and reproducibility prevails, using open-access primer sequences is preferred. Relying on companies for consistency in their product QC can be tricky and a minefield for data reproduction. Here is a paper providing a TRIC FWD and REV primer sequences (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1698716/)
      These were the bad part in my opinion, now there are also good part and there are some interesting findings that would be interesting if you sorted the data by age-match, ethnicity or gender to see if there is interesting stuff into these data. It would be also interesting to see how your RNA data results in changes in protein levels. Especially on how MRI scans from ASD look like from controls (are they showing some imaging features indicative of BBB leakiness?). My own little bias is that maybe differences in transporters at the BBB may matter more than TJ complexes, like the BCAA transporter recently described by Novarino and colleagues (https://www.ncbi.nlm.nih.gov/pubmed/27912058). I am very intrigued if the gut-brain axis in ASD maybe due to impaired metabolites rather than structural changes. Remember the BBB is here for a reason and even during stroke, you don’t have a “breakdown” like a dam that get breached because you would be dead. It is more subtle, more discrete and involve a BBB remodeling with all the tutti-quanti following stroke (second BBB opening during 24-48 hours post-stroke, endothelial cell activation, loosening of the ECMs at the BM….). Thats my own personal bias.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s