APOE4 increases inflammation related pathways
July 18th, 2025 By John Widen
An open access article focused on how ApoE4 affects protein levels in humans was published this week in
Nature Medicine by a group in the Westmead Institute for Medical Research
at the University of Sydney, Australia. The group relied on a technology called SomaScan, which utilizes aptamers to measure thousands of proteins
in samples simultaneously by tagging proteins with unique DNA sequences. According to this publication, it can detect 7,000 different proteins in samples
out of the estimated 22,000 proteins in the human proteome. A quick internet search shows that since this publication the
company SomaLogic can now detect 11,000 proteins
simultaneously. They didn’t pay me to link to their website and I didn’t dive any deeper than that because I want to focus on this publication.
The team measured protein levels in human CSF (cerebral spinal fluid), plasma, and post-mortem dorsolateral
prefrontal cortex from combined cohorts across the United States, UK, and Europe. They measured a grand total of
11,270 CSF and plasma samples from the Global Neurodegeneration Proteomics Consortium (GNPC). That is A LOT of samples!
Whoa! It is impressive to obtain and test that many real-world samples. In addition, they evaluated 262 most-mortem samples
using label-free MS proteomics assays from donors with various cognitive diseases. Again, VERY impressive data set.
I’m guessing this is how it made it into Nature Medicine.
All of these samples were from donors with various cognitive diseases including Alzheimer's disease, frontotemporal dementia (FTD),
Parkinson’s disease dementia (PDD), Parkinson’s disease (PD), and amyotrophic lateral sclerosis (ALS) as well as nonimpaired controls.
The group argues that APOE4 is not just a risk factor for Alzheimer’s but can have effects across many different
disease types (i.e. pleiotropic effects), which seems to be likely based on their data.
I was very excited when I started to read this publication but that quickly devolved into a general lackluster feeling, unfortunately.
The overall message from this publication is that ApoE4 causes mild increases in inflammatory
pathways in the brain and periphery.
After likely spending millions of dollars and using thousands of human samples on analysis,
this study came to a conclusion that was already well established in the neurodegeneration field.
At least it agrees with the great body of evidence that exists in the literature pointing towards ApoE4
contributing to neuroinflammation.
However, their analysis is a bit odd from my perspective. For some reason, they only looked for differences in protein levels
between APOE4 carriers and APOE2&3 carriers in non-impaired individuals. Maybe this was because they were scraping the
bottom of barrel for something to publish about. Notably, the paper also alludes to a lot of non-correlations to Abeta plaque and
tau aggregate levels, which is interesting information but not surprising. Alternatively, they could be gating data for
a separate publication or holding onto it for commercialization.
After downloading the supplemental information and looking through the protein lists, I didn’t find anything particularly exciting.
The curated lists of proteins from each sample set doesn’t provide p-values, how many samples the proteins were found in, raw or normalized
signals from the assay, or any other information regarding the protein levels. Maybe because SomaScan is proprietary. The p-values can be
estimated from volcano plots in the extended information.
They chose to focus on a statistical analysis called mutual information, which to my understanding is a measure of how two datapoints
are related to one another. In this case, how well a protein of interest can predict whether the sample is from an ApoE4 carrier or not. In my humble
opinion, this analysis is not the most useful because APOE4 is a RISK factor, not sufficient for Alzheimer’s disease. There is already
a cheap and easy genetic method to know whether you have this risk or not, if anyone wants to find out. I don’t think running an expensive
proteomic analysis on blood plasma is necessary. They did not discuss other useful information these data sets could be used for such as stratifying
patient populations. But again these were changes between ApoE4 carriers and non-carriers in healthy individuals.
The study didn not find many differences in protein levels and the differences by their mutual information analysis seems weak and sometimes
confounding between sample types.
In CSF there were 229 statistically different proteins between ApoE4 carriers and non-carriers. I think this is likely because they were only
looking at non-impaired samples, which again is a weird choice. Additionally, they didn’t differentiate between heterozygous and homozygous APOE4
carriers, which as most know have very different odds ratios (3 vs 31 respectively) for Alzheimer’s disease
(Reiman Et al. Nature Comm. 2020 11, 667).
Maybe I’m off, but I would think that all of these confounding factors would wash out meaningful differences between these sample sets. Another limitation
is that they are only detecting less than one third of the proteome at maximum efficiency, which also hinders the overall analysis.
The protein lists contain many highly abundant proteins such as structure related proteins and chaperones.
Their pathway analysis, which was the crux of the publication, identified immune related pathways. However, there are many key
protein classes missing that I would expect to find such as interleukins and other cytokines related to inflammation or immune related pathways
involving Jak/Stat, Interferon Response Factors (IRFs), or NF-kappaB. There are some proteins involved in these pathways but I would expect a
stronger showing with such a rich data set. Feel free to look through the list yourself and let me know what you think. I’m sure there are interesting
proteins in these lists for additional studies.
The last thing I will discuss is the relevance of these differenitated proteins. From a medicinal chemistry perspective there does not seem to be
anything to follow up on in regards to actionable protein targets for therapeutic intervention. The pathways revealed from this analysis, as mentioned
previously, are already known to be involved in neurodegeneration. Many of which are being targeted for therapuetic intervention. I hate to be negative
in one of my first articles and I hope that there is more to come from this large data set.
The site does not have a comments section yet! Hopefully, very soon! Until then please drop me a line at jwiden@chemjam.com.
If you provide comments on my articles I reserve the right to post them on this website as additional commentary. My goal is to have an open discussion!
