[Neurosciences/Aluminum] Does the latest paper from Exley show a link between ASD and aluminum?

Someone brought my attention today about the most recent Exley paper out in the press titled “Aluminium in brain tissue in autism” (the title could have been better but well….) and published in the journal “Journal of Trace Elements in Medicine and Biology“.
Let me put this straight, this is not a paper that has evidence of scientific fraud or data manipulation. There is no duplicated images, no suspicious blots. The problem I have with this paper is its deep experimental flaws and data analysis that nonetheless should not have passed through the peer-review filter.

1. Before we dive into the paper, lets put the paper into context
   Lets just put the paper in the context. It was received on October 26th (Thursday). Came back in its revised form on November 21st on Tuesday and accepted for publication on November 23rd (Thanksgivings for the US, but since the editor-in-chief (EIC) is in Europe no Thanksgiving here). Let that sink it a bit: in a bit more than three weeks, it got send to review, came back from review and got revised in 26 days. In my standard of reviewing for journals and publishing my papers, thats some faster-than-light peer-reviews. I usually wait 4-5 weeks by the time I submit mine and get the editor reply to my submission with the infamous reviewers comments. Does a fast-reviewed manuscript means a bad manuscript? Not necessarily, but it can mean that maybe the peer-reviewed was not optimal, rushed or even worse just botched. Based on the quality of the data presented, I am leaning towards a botched review. Thats quite disappointing because the journal holds a decent impact factor (~3 for 5-year impact factor) and you expect an okay review.
   Then comes another problem. Exley published this paper (as well as few others) in the journal…..in which he holds a seat in the editorial board. Nobody can exclude the possible conflict of interest. Consider that: if you were an EIC, would you provide the same rigor and objective decision on a paper submitted by a colleague sitting in your editorial board than a paper submitted by Doe and colleagues?
   Not forbidden, but if you can avoid it, avoid it. Transparency is key and publishing in diversified journals (unless it is society-official journals) is an indicator of an healthy research.
   Finally, the last thing to keep in mind before I deconstruct the paper is the funding source. According to the acknowledgment section “The research is supported by a grant from the Children’s Medical Safety Research Institute (CMSRI), a not-for-profit research foundation based in Washington DC, USA.”  Behind the fancy name is just another anti-vaccine foundation that will play “the vaccine safety” card to peddle their pseudosciences. So we can claim that Exley is a shill for CMSRI, since he received monetary support for his research. Does that mean the research is completely bogus? No, but it means it will require further scrutiny, especially when the claim of the study goes against the consensus in the field (aluminum in vaccines is safe).
   Same goes if a study funded by Big Tobacco claimed the absence of correlation between lung cancer and smoking or if Big Sugar claimed the absence of correlation between type 2 diabetes mellitus and consumption of sweetened beverages.
2. So what is wrong with this paper?
   For those who wants to read the paper with me, you can download it here (I assume it is open-access, so you should not have an issue with the paywall). Exley has a publication record on aluminum, especially when it comes to its possible ecotoxicity and the impact of aluminum on certain biological processes.
   The introduction is damn short, half a page of a double-spaced document but set the tone, this study will investigate the relationship between autism and aluminum in the brain.
   Samples are obtained from the Oxford Brain Bank, but felt short to indicate the source of the tissue (like a catalog number) and how this source of materials was complying with the institutional review boards (IRB). Basically, for any research involving human subjects or human tissues, you have to comply with the IRB that such specimens are used for a certain and defined use and foremost been anonymized.
   We have 5 patients that were diagnosed as on the autism spectrum and immediately we can pinpoint an important issue: there are no controls and that’s one of the big and unforgiving flaw of this paper.
   The authors then used two techniques to localize and quantify the Al in different cortical regions (and sometimes hippocampal regions). They have used three technical replicates (random sampling from the same cortical lobe) for measuring the Al content using an atomic absorption spectrometry and used lumogallion (aka4-chloro-3-(2,4-dihydroxyphenylazo)-2-hydroxybenzene-1-sulphonic acid, a fluorescent dye initially described to localize Al in plant roots). This dye have an excitation/emission spectra close from FITC/Alexa Fluor 488. It has been also used for live cell imaging , in particular to study how macrophages process Al present in vaccines adjuvants (http://www.sciencedirect.com/science/article/pii/S0022175915001222).
   Considering the equipment mentioned in the method, the microscope used provides the right excitation bandwidth filter and provide a long pass emission filter for anything over 510nm.Then things get weird, in the result sections, the authors mention the following:”We examined serial brain sections from 10 individuals (3 females and 7 males) who died with a diagnosis of ASD and recorded the presence of aluminium in these tissues (Table S1).“Where is the number coming from? Why don’t we have the same numbers in the Materials and methods?The other problem is the over interpretation of the data. To be brief, the lumogallion will show some punctuated pictures. The authors show some brightfield pictures overlapping to show the tissue structure but does not really help the reader. A DAPI stain (to stain cell nuclei) as counterstain would have been much more informative, it would helped to distinguish background noise from possible Al inclusion. Again, keep in mind we have no controls. The other issues with immunostaining is the high risk to cherry pick the data. You will be naturally inclined to show the presence of a positive risk but this cannot be used for quantitation. Thus, the use of the second method is welcomed as a complementary technique.
   For those not familiar with fluorescence, there is an important notion to keep in mind when analyzing the data: ensuring you keep the same exposure time, the same brightness or contrast and foremost have a negative control to set your exposure time. You can see a sketch explaining here on one of my fluorescence staining (based on my data, I concluded the expression was weak if not negative).

   

   The background subtraction is also a bit weird. I acknowledge the assessment of autofluorescence is a good control, but you expect to see a low staining. But foremost, you cannot overlap two distinct slices, as proximal as it can be. For instance, in Figure 1, you see some lumogallion staining and below the fluorescence  from the “control” using the adjacent slice. The lumogallion also seems to have a very high background.
   
   It seems lipid-rich environment increase dramatically the fluorescence of lumogallion (if you look at the spectra, the dissolution of the dye in Triton-X100 solution (b, a detergent) dramatically increase the excitation and emission spectra compared to water (a)).
   What I found troubling is this sentence in the results section: “We examined serial brain sections from 10 individuals (3 females and 7 males) who died with a diagnosis of ASD and recorded the presence of aluminium in these tissues (Table S1). Excitation of the complex of aluminium and lumogallion emits characteristic orange fluorescence that appears increasingly bright yellow at higher fluorescence intensities. Aluminium, identified as lumogallion-reactive deposits, was recorded in at least one tissue in all 10 individuals. Autofluorescence of immediately adjacent serial sections confirmed“.
   If you are a bit a fluorescence microscopy savvy, you know that the “emission color” we see in the objective is never caught by the CCD camera. These camera have in the most majority a B&W output for the simple reason that they have a much higher sensitivity than color cameras. You can always re-create colors in the micrograph pictures using various “lookup tables” (LUTs) that will give a pseudo color based on the level of grays. This is very useful when you samples different excitation/emission channels (for instance, samples stained with DAPI and two antibodies, one conjugated with Alexa Fluor 488 and the other with Alexa Fluor 546 or further down).
   The problem inherent with fluorescence is you can make thing fluoresce or end up with a false-positive signal if you increase the light beam (usually never happens because it is set) or if you increase the exposure time of your camera (this is the most common issue). As you increase exposure, you increase the risk to capture non-specific signal like autofluorescence signals.
   The other problem here is how to explain this sudden shift from orange to yellow?  This seems more like a subjective observation than something caught on camera.  That can be due to different things. You can have some bleed-through of the dye that is normally emitting in a certain wavelength but if it is strong enough can appears in neighboring emission channels. This thing rarely happens with a good fluorescence microscope that have defined filter cubes that allows the diffusion of certain emission wavelengths (for instance, my microscope have a DAPI, Alexa Fluor 488 and Alexa Fluor 555 cubes that only let the respective emission wavelengths  with 20nm-margin error to cross through the objective and reach the camera and binocular).
   Usually, we have to deal with bleed-through when you use flow cytometry and usually is solved using fluorescent dyes latex beads and by following a protocol called “compensation” (this has the result of removing any noise and keeping only the signals).
   We cannot also exclude that such fluorescence is just an autofluorescence from lipofuscine inclusion bodies. Lipofuscin is a lipid-based compound naturally produced by our cells. It has an important concentration in the central nervous system, however it is normally cleared out by cells. Failure in the clearance of lipofuscin is associated with different diseases called “lipofucsinosis” such as Batten’s disease. Even the author admit the possible presence of lipofuscin inclusions “Intracellular aluminium was identified in likely neurones and glia-like cells and often in the vicinity of or colocalised with lipofuscin (Fig. 5).” Lipofuscin is also capable of autofluorescence, although it is more in the wavelengths matching DAPI. Lipofuscin has an excitation/emission peaks at 360 and 435nm respectively but has been reported to also show fluorescence at 510nm when excited at 488nm (https://www.sciencedirect.com/topics/neuroscience/lipofuscin).
   Compared to the lumogallion excitation/emission spectra (507/567), we cannot exclude the presence of a phenomenon called “FRET” (Fosterman Resonance Energy Transfer) in which the excitation of lipofuscin (as the microscope excitation bandwidth is 470-495nm) provide enough energy to the photons emitted by the lipofucsin to excite nearby lumogallion dyes. Because the microscope setting used in this paper has no restricted bandwidth (it let pass any photons harboring a wavelength of 510nm and more), it may explain this orange-to-yellow transition noted by the author. The presence of a DAPI nuclear stain would greatly helped to identify this region as grey matter (rich in cells) or white matter (rich in lipid-rich myelin sheets). Thus, we can legitimately questions the nature of these as it these punctae labelled as “Al inclusion” are simply lipid inclusion or some artificial noise due to the tissue processing. This is where controls come as critical, it can help you sort the signal from the noise.

    

   The second big issue with this paper is the over-interpretation of what the experimenter see. The experimenter wants to see Al inclusion in monocytes? So be it: “Aluminium-loaded mononuclear white blood cells, probably lymphocytes, were identified in the meninges and possibly in the process of entering brain tissue from the lymphatic system“. Or maybe these are astrocytes, or neurons, or microglial cells, or blood vessels….or whatever the author wants to believe in: “Aluminium could be clearly seen inside cells as either discrete punctate deposits or as bright yellow fluorescence. Aluminium was located in inflammatory cells associated with the vasculature (Fig. 2). In one case what looks like an aluminium-loaded lymphocyte or monocyte was noted within a blood vessel lumen surrounded by red blood cells while another probable lymphocyte showing intense yellow fluorescence was noted in the adventitia (Fig. 2b). Glial cells including microglia-like cells that showed positive aluminium fluorescence were often observed in brain tissue in the vicinity of aluminium-stained extracellular deposits (Figs. 3&4). Discrete deposits of aluminium approximately 1m in diameter were clearly visible in both round and amoeboid glial cell bodies (e.g. Fig. 3b). Intracellular aluminium was identified in likely neurones and glia-like cells and often in the vicinity of or colocalised with lipofuscin (Fig. 5). Aluminium-selective fluorescence microscopy was successful in identifying aluminium in extracellular and intracellular locations in neurones and non-neuronal cells and across all brain tissues studied (Figs.1-5). The method only identifies aluminium as evidenced by large areas of brain tissue without any characteristic aluminium-positive fluorescence (Fig. S1).”
   This is the second big mistake of this paper. If the author wants to make the claim he proposed here, then he has the obligation to show a counterstain using selective markers for neurons (e.g. MAP2, bIII-tubulin, NeuN….), astrocytes (e.g. GFAP), microglial cells (CD11b), leukocytes (CD3), macrophages (CD45), blood vessels (e.g. PECAM-1, claudin-5). This could have been easily performed (using a secondary antibody conjugated with Alexa Fluor 555 or better Alexa Fluor 647)  and would have give support to this claim.
   If the author can identify cells by the naked eye, he is either equipped with  Superman X-ray eyes or he is just imagining things.

   The discussion quickly gets into an anti-vaxxer diatribe and throws the minimal amount of scientific data under the bus.
   For example, the author throws this sentence as is: “We recorded some of the highest values for brain aluminium content ever measured in healthy or diseased tissues in these male ASD donors including values of 17.10, 18.57 and 22.11 g/g dry wt. (Table 1).” Firstly, where does it get this data? You cannot sum technical replicates, you have to average them (even with considering the huge variability between technical replicates). Secondly, how can the author make a claim like this without providing values from controls (well there are no controls) or from the literature. It is like “we have recorded the highest amount of leukocytes in ASD patients blood samples with values of 11.3, 12.0 and 11.5 x10e3 cells/mm3.” I cannot make an interpretation or conclusion without knowing the reference from the normal population (normal range 4.5-11x 10e3 cells/mm3) or from control groups. The average Al level was 2.38-4.79 microg/g tissues in male ASD and 1.15 in the female ASD patient. Such levels were very similar to those reported in samples from patients suffering from familial form of Alzheimer’s disease.
   

   The data is interesting but we are lacking additional female samples to make a claim as he did: “All 4 male donors had significantly higher concentrations of brain aluminium than the single female donor.” He lacks the proper conditions to run the statistics (you need same number of patients in male and female to make such claims) and even the important inter-individual variability makes it unlikely that he could achieve the statistical significance. This is a statement that would put a graduate student in shame for overconfidence in the data.
   Then goes the tirade “What discriminates these data from other analyses of brain aluminium in other diseases is the age of the ASD donors. Why, for example would a 15 year old boy have such a high content of aluminium in their brain tissues? There are no comparative data in the scientific literature, the closest being similarly high data for a 42 year old male with familial Alzheimer’s disease (fAD) [19].” (another Exley paper published…..in the same journal). We are dealing with the same issues (lack of controls, huge variability in the technical replicates…..).
   Now if you plot the average patient Al levels agains the age, regardless of the condition, you end up with an homogenous cloud. Now, two things have to be noted here: seems there is no impact of Al levels based on the disease (only age seems to matter between ASD and AD) and there is no correlation between increase in brain Al and age, at least in the very small sample size.
   

   No pun intended, but the data scatter looks vaguely like the United States map. Again, it shows the need of data from asymptomatic patients to estimate the burden of Al in the brain.
   Since we have not access to Al content in the brain, we have to see some values in the literature. A study by Andrasi and colleagues (https://content.iospress.com/articles/journal-of-alzheimers-disease/jad00432) provide some Al levels in control samples. According to their study, the average Al content in control samples were between 1.4 to 2.5µg/g dry tissue. We are indeed not far from the value reported by this study, especially when you consider the important standard deviation in these samples.

   Maybe it is also to consider the other study by Exler on Al level in brain samples from patients associated with familial form of Alzheimers disease (fAD) and familial dementia. In that study, all reported with Alzheimers (some with early onset, some with late onset based on age), the Al values reported were ranging from 0.34microg/g tissue (male) to 6.55microg/g (female, presenting a mutation in the PSEN1 gene, a known gene in FAD). So are we just measuring noise and try to extrapolate data from noise? Thats some bold statement that should have been smashed already by a decent reviewer in the field of neurosciences.
   But seeing these two papers went through in a apparent free ride is not looking good for the journal integrity.

3. Conclusive Remarks
   To make a claim is one thing, to back it up with robust data is another thing. I think Exley jumped the shark a while ago and started to aluminum as the big bad wolf in every little things. But a wolf can be tamed, kept out from showing danger to the community and somehow co-exist. But for Exley, like Shaw, like Gherardi, aluminum is the devil incarnate. God forbid it has been used for 70 years and showed barely more than simple coincidence in its association with some disease, aluminum is their dead horse that worth being beaten again and again. If your funding sponsor will give you money for showing a link between aluminum and autism, lets give them what they want. Ethically it is insane, but when you need to keep your lab and your faculty position afloat, sometimes making the pact with the devil and throwing the scientific integrity and the philosophism that is given to you  following your thesis defense can be tempting. Sometimes, it feels that anti-vaccines researchers are like Faust and succumbed to the offer made by Mephistopheles offer. But this come with a price and a hefty price to pay: the loss of your integrity as a scientist.
   So my question is what is coming next to patients on the spectrum: does this study will be used to support the anti-vaccine agenda (another reason to yell “Aluminum is a chemikillz” in parenting groups?) and breakdown the herd immunity? Bogus remedies by bleach enemas and drops (the infamous CD/MMS)? or give a support to chelation therapy? gluten-free/casein-free diet? Or like Exley once claimed have these people drink ad nauseam silicon-rich water like Fiji water or Volvic water with the magic claims that the silicon with drain your brain from the Al contained inside it?
   This kind of deeply-flawed studies, lacking proper controls and driven by an ideology over the facts are dangerous because they prey on the meek and enrich modern snake oil sellers.