Preprint Charting the Circulating Proteome in ME/CFS: Cross System Profiling and Mechanistic insights, 2025, Hoel, Fluge, Mella+

[Hutan]

2. The study is yet another powerful indicator that there is no inflammation or damage going on.

I understand the point you have made previously about inflammation being a specific thing with heat and swelling and white blood cells, and that not seeming to be part of ME/CFS. And about the importance of being specific about what is meant when inflammation is mentioned.

But, I'm not sure what you are seeing in this particular paper that supports the absence of inflammation. The authors seem quite convinced that they are in fact seeing evidence for inflammation. Perhaps it is just that different idea about what inflammation is?

Searching on 'inflammation' in the paper produces content like these quotes:

Conclusion - Our findings characterize a pathophysiology in ME/CFS composed of interrelated elements, including immune dysregulation, vascular dysfunction, metabolic stress, and chronic inflammation. The observed multiscale patterns of change converge on a mechanism involving coordinated immune, vascular, and metabolic disturbances.

In summary, the pattern of elevated secretory activity in the ME/CFS group is mechanistically consistent with immune dysregulation, inflammatory responses, and metabolic stress.

This clearly showed a skewed pattern of increase for affected secretome proteins involved in coagulation, complement pathways, and inflammation (chemokines, interleukins).

Immune dysregulation, as reported in multiple studies 15,16,55, may be mechanistically associated with an underlying autoimmune pathomechanism, the overall state of elevated secretory activity, and increased levels of factors involved in inflammation, coagulation, and complement activity. Similar observations have previously been highlighted in relation to ME/CFS and long COVID

For the next two quotes, I'm not really sure if they are saying the proteins increase or decrease inflammation:

To specifically address aspects of immune dysregulation, we used annotated protein panels (HPA) specific to B-cells, T-cells, NK-cells, dendritic cells, monocytes, and granulocytes (Figure 4D). The effects associated with these cell types pointed to possible impacts on immune cell interactions, coagulation, and inflammation. As previously mentioned, multiple proteins associated with granulocytes showed significantly lower levels in the ME/CFS group compared to the HC group, such as MPO and BPI. The reduced amount of granulocyte proteins was not associated with abnormally low neutrophil counts (the most abundant type of granulocytes) or other leukocyte types in the patients (DocumentS1, Table S1). Furthermore, comparing a list of proteins associated with neutrophil granules and stimulated neutrophil protein release 33 45, we found that about 40% or more of the proteins reported to be released by activated neutrophils showed lower serum concentrations in the ME/CFS group compared to the HC group, suggesting a suppressive effect on overall neutrophil activity.

We performed a ligand-receptor interaction analysis to elucidate possible patterns of regulation that could be related to the pathomechanism of ME/CFS (DocumentS1, Figure S1). Overall, we found an overweight of cell adhesion and cytokine-cytokine receptor type ligand-receptor interactions. Interestingly, four members of the Ephrin subfamily A receptors (EPHA) and their ligands (EFNA) displayed concordance, supporting the involvement of altered EPHA-EFNA signaling in ME/CFS, as previously suggested 29. Furthermore, there was a positive concordance between Fibroblast growth factor receptor 3 (FGFR3) and EPHA, IL6R with the CNTF and MPZ ligands, and FLRT3 with UNC5B and UNC5D. FAP and PAM had negative concordance, along with other interactions. Based on current knowledge, it seems likely that the identified ligand-receptor interactions may play a role in ME/CFS pathology through their impact on metabolic regulation, tissue development and repair, inflammation, and angiogenesis.

 

[Jonathan Edwards]

But, I'm not sure what you are seeing in this particular paper that supports the absence of inflammation. The authors seem quite convinced that they are in fact seeing evidence for inflammation. Perhaps it is just that different idea about what inflammation is?

There are a shedload of things that go up with inflammation that are not here. All the acute phase proteins for instance.

As far as I can see the discussion in the paper simply reflects "well the keys must be here somewhere under the lamp-post because that is where I have to look". The conclusion statement is so vague it could mean anything.

Most people writing scientific papers have an irresistible urge to go on saying what everyone else is saying. But of course progress in science comes from the opposite.

 

[butter.]

Another group is showing that GDF15 is normal in ME/CFS, @Jonathan Edwards. The testing I was part of showed that the most severe patients had the lowest GDF15. Explaining how this could be would go a long way, in my opinion.

I still think it's plausible that, due to poor microvascular diffusion and reduced gradients, blood values might not reflect tissue values. Something similar is seen in mitochondrial diseases, where lactate levels in the blood can be normal.

There could be many other reasons, of course.

 

[mariovitali]

Just realised that among the downregulated proteins is IFI16, which is Gamma-interferon-inducible protein Ifi-16. Not sure if this is relevant to the theory of @Jonathan Edwards so making sure this is noted.

https://en.wikipedia.org/wiki/IFI16

 

[Jonathan Edwards]

I still think it's plausible that, due to poor microvascular diffusion and reduced gradients, blood values might not reflect tissue values. Something similar is seen in mitochondrial diseases, where lactate levels in the blood can be normal.

Lactate diffuses back across venues to the bloodstream but proteins go back via lymphatics and I don't see any reason to think lymphatics are a problem.

 

[Jonathan Edwards]

Just realised that among the downregulated proteins is IFI16, which is Gamma-interferon-inducible protein Ifi-16. Not sure if this is relevant to the theory of @Jonathan Edwards so making sure this is noted.

My prediction seeing this was that IFI16 would be down regulated by TGFbeta, so would show a compensatory fall.

It is down regulated by TGFbeta!

 

[butter.]

Lactate diffuses back across venues to the bloodstream but proteins go back via lymphatics and I don't see any reason to think lymphatics are a problem.

But isn't the (capillary) lymphatic uptake of proteins itself dependent, at least indirectly, on a properly functioning microvasculature? Also, wouldn't cells that aren't adequately taking up oxygen produce relatively fewer proteins?

I guess, I am trying to understand how GDF15 is normal in pwME? Isn't it curious that it's normal in such severely incapacitated patients? I'm not sure whether the lower GDF15 levels in more severe patients represent a real trend, but if they do, that’s even more puzzling.

Then another thing re lymphatics, anecdotally, my massage therapists have often mentioned something like a "palpable buildup of lymph" in my tissues. I'm not sure whether that makes any real sense. It feels like small nodules, and when pressure is applied, it almost feels like something "pops."

 

[Kitty]

But, if there is an allele that increases interferon gamma responsiveness to a signal, perhaps that could be a reason for the female predominance in ME/CFS?

I had the same thought yesterday. Diseases where the susceptibility is on the X chromosome are more likely to affect men because they (mostly) only have one X. But if it was an enriching effect rather than loss of function, it could affect women more.

Didn't post it because I decided I'd no idea what I was talking about.

 

[ME/CFS Skeptic]

For what it is worth Leptin was also increased in ME/CFS patients in this group. Here's my own plot of the raw data (I used Aptname seq.2575.5)




The authors do note that much of the variance in leptin was explained by BMI and sex:

 

[forestglip]

The authors do note that much of the variance in leptin was explained by BMI and sex:

I was looking at that, wondering if BMI was the reason for the increase (matched for sex), but they seem pretty closely matched for BMI as well:

HC: 24.0±2.6
ME/CFS: 24.4±4.2

Maybe the larger variance in BMI in the ME/CFS group could affect things?

Edit: Oh, not perfectly matched for sex like I had assumed. It says 72% are women in the HC group and 80% in the ME/CFS group. Maybe it'd be worth testing while controlling for these two variables.

Edit: And it says "HC BMI N=24" so I guess 5 of the HC didn't have BMI data.

 

[ME/CFS Skeptic]

Maybe the larger variance in BMI in the ME/CFS group could affect things?

The fact that age and BMI have a large influence on leptin levels, doesn't necessarily exclude that ME/CFS has an additional effect on top of those factors I suppose (it does probably mean that matching of groups is really important). It's notable that it keeps coming back in different studies, even in those where BMI was appropriately matched like this one.

In the Hanson study that used aptamers (Germain et al. 2021), the p-value was not significant (they only had 20 patients) but the fold change changes suggests it was higher in patients versus controls.

 

[ME/CFS Skeptic]

Was looking at how the results of this Norwegian study (Hoel et al. 2025) could be compared to the aptamers study from the Hanson group (Germain et al. 2021) but ran into a problem of not having a good effect size to compare.

The fold change simply gives a ratio of the abundance of the protein in both groups without taking variability among participants into account. So a large fold change doesn't mean big effect because it might be explain by high variability among participants.

The p-value isn't ideal either because it is dependent on the sample size.

 

[ME/CFS Skeptic]

Was looking at how the results of this Norwegian study (Hoel et al. 2025) could be compared to the aptamers study from the Hanson group (Germain et al. 2021) but ran into a problem of not having a good effect size to compare.

Perhaps we could first filter based on small p-values in the Norwegian study and then compare how the fold changes of these proteins relate to those the Hanson study.

 

[ME/CFS Skeptic]

Perhaps we could first filter based on small p-values in the Norwegian study and then compare how the fold changes of these proteins relate to those the Hanson study.

Had a go using this method but I'm far from the best coder so might have made a mistake. I got a relatively weak correlation of 0.33. In the Hoel et al. dataset there were 845 observations with an adjusted p-value below 0.05 and after an inner join with the Germain et al. dataset there were 672 observations left.

EDIT: given that these are ratios I shouldn't have probably calculated a correlation on it (they are not symmetric: 1/3 = 0.33 while 3/1 = 3), so I have now used the log fold change.

 

[ME/CFS Skeptic]

Attempt to highlight some proteins that were increased in both datasets (leptin appears twice because it was measured twice in the Hoel study):

 

[forestglip]

Had a go using this method but I'm far from the best coder so might have made a mistake. I got a relatively weak correlation of 0.3. In the Hoel et al. dataset there were 845 observations with an adjusted p-value below 0.05 and after an inner join with the Germain et al. dataset there were 672 observations left.

I'm running into the issue when merging with an inner join of not having a consistent identifier. If I merge on 'UniProt', I get 682 rows. If I merge on 'EntrezGeneSymbol' I get 655. If I merge on 'EntrezGeneID', I get 631. Not sure how yours is different from all of these.

It's different between identifiers because of things like where the EntrezGeneSymbol is 'NRXN1', one study has UniProt as 'P58400' while the other has 'Q9ULB1'. Where they both have EntrezGeneSymbol of CBS, one has EntrezGeneID of '102724560', the other has '875'. Where they both have EntrezGeneID of '23526', one has EntrezGeneSymbol of 'ARHGAP45', the other has 'HMHA1'.

 

[ME/CFS Skeptic]

I'm running into the issue when merging with an inner join of not having a consistent identifier. If I merge on 'UniProt', I get 682 rows. If I merge on 'EntrezGeneSymbol' I get 655. If I merge on 'EntrezGeneID', I get 631. Not sure how yours is different from all of these.

I used an inner join with as key 'TargetFullName'. It's possible that some proteins were lost because their name wasn't identical in both datasets.
EDIT: perhaps we should just use the one that gives the highest amounts of rows?

 

[forestglip]

I used an inner join with as key 'TargetFullName'. It's possible that some proteins were lost because their name wasn't identical in both datasets.

Oh, another identifier, didn't notice that one. I do get 672 with that. I'm going to try merging where any of the columns match.

Could you explain this a bit more:

EDIT: given that these are ratios I shouldn't have probably calculated a correlation on it (they are not symmetric: 1/3 = 0.33 while 3/1 = 3), so I have now used the log fold change.

 

[ME/CFS Skeptic]

Could you explain this a bit more:

The fold changes are not symmetric around 1. It's a ratio of the abundance of the ME/CFS group versus controls.

So if the ME/CFS group had higher values, the ratio could be somewhere between 1 and infinity.
If the ME/CFS group had lower values it would be somewhere between 0 and 1.

That makes it problematic to calculate a correlation. Using log2 fold changes makes it symmetric. A logFC of +1 means the protein is 2× more abundant in the case group; a logFC of −1 means it's 2× less abundant.

 

[forestglip]

EDIT: perhaps we should just use the one that gives the highest amounts of rows?

I think TargetFullName might be best actually. Multiple aptamers can be associated with the same gene identifier, but I assume they each have a unique TargetFullName. If matching on UniProt, if either dataset has multiple aptamers per UniProt ID, you'll get arbitrary pairs of measurements between studies likely corresponding to different aptamers.

I think each row in Germain is an aptamer, so it would have been great if they had included the aptamer identifier as a column like Hoel so they could be matched, but TargetFullName might be a decent proxy. (Though I don't know for sure each aptamer corresponds to a unique TargetFullName and vice versa.)

Edit: Nevermind, they're not unique. Multiple aptamer IDs match to the same TargetFullName in the Hoel data. So I'm not sure there's any good way to match them up. Maybe the fold change in each study can be averaged for rows with matching TargetFullName before merging, so there's only one of each name in each dataset.