Preprint Transfer of IgG from Long COVID patients induces symptomology in mice, 2024, Vidarsson+

Transfer of IgG from Long COVID patients induces symptomology in mice
Vidarsson, G., Bos, A., van Vugt, M., den Dunnen, J., Wiersinga, W. J., Wust, R., Amsterdam UMC COVID-19 biobank,, Versteeg, S., Silva Santos Ribeiro, P., Willemen, H., Geyer, C. E., Larsen, M. D., Prado, J., Charlton, B., Schuchner, E., Bomers, M. M. K., Appelman, B., Eijkelkamp, N., Lavell, A. H. A., Chen, H.-J.

SARS-CoV-2 infections worldwide led to a surge in cases of Long COVID, a post-infectious syndrome. It has been hypothesized that autoantibodies play a crucial role in the development of Long COVID and other syndromes, such as fibromyalgia and myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS).

In this study, we tested this hypothesis by passively transferring total IgG from Long COVID patients to mice. Using Glial Fibrillary Acidic Protein (GFAP) and type-I interferon expression, we stratified patients into three Long COVID subgroups, each with unique plasma proteome signatures.

Remarkably, IgG transfer from the two subgroups, which are characterized by higher plasma levels of neuronal proteins and leukocyte activation markers, induced pronounced and persistent sensory hypersensitivity with distinct kinetics. Conversely, IgG transfer from the third subgroup, which are characterized by enriched skeletal and cardiac muscle proteome profiles, reduced locomotor activity in mice without affecting their motor coordination.

These findings demonstrate that transfer of IgG from Long COVID patients to mice replicates disease symptoms, underscoring IgG's causative role in Long COVID pathogenesis. This work proposes a murine model that mirrors Long COVID's pathophysiological mechanisms, which may be used as a tool for screening and developing targeted therapeutics.

Link (Biorxiv)

 

See also Students investigate effect of immune system of Long COVID patients on muscle cell function as well as the presentation "Transfer of IgG of Long-COVID patients induces subgroup specific symptoms in mice" from a recent conference.

Note also that this group has supposedly registered for a patent for a Long-Covid biomarker on the basis of this work and that they have funding for different follow-up projects, including one on ME/CFS.

 

This is not the first time GFAP has appeared in the LC literature (see for instance Neurofilament light chain and glial fibrillary acid protein levels are elevated in post-mild COVID-19 or asymptomatic SARS-CoV-2 cases and the numerous references listed there). It raises the question whether there is something to those findings, whether it is just related to acute Covid and/or whether GFAP measurements are simply something that tend to be less robust and consistent and as such noise is more commonly picked up, especially if GFAP values are more prone to change due to Covid infections.

Unfortunately, from what I’ve seen the authors don’t seem to be reporting the actual values or the p-values and upon quickly glimpsing over the paper I didn’t see how those values are corrected for when accounting for multiplicities (they do have some description in the "Statistical Analysis" section, but it's hard to quickly grasp for me what applied to what).

 

It's very much appreciated that the authors decided on submitting a preprint.

I can't claim to have fully understood the paper yet.

The authors did a lot of different experiments across different time points. The different time points were analysed via a linear mixed effects model, which is vaguely similar, but used in different situations, to how the intramural study used GEE to analyse the EEfRT data to account for different interactions across multiple rounds of the EEfRT. They corrected for multiple comparisons using Benjamini-Hochberg. Unfortunately, since the actual data isn't part of the paper all one has is the statements of the authors on certain things turning out statistically significant, without actual corrected p-values being reported anywhere, although given the amount of analysis necessary to perform, most readers in any case could only do very little with the actual data (at least I would have no idea what to do with it, but I'd at least like it to be there for others to verify and for future purposes as well as other researchers working on replication).

In the mice experiments supposedly the authors used 8 different HC's and 8 different people from each of the LC groups. They didn't specify this, but anything else wouldn't make sense (but it could be interesting to use the same person on two different mice to see if those results are consistent). They only report the group differences for the mice so it's not possible to tell whether the differences are being driven by outliers, either being certain patients or being certain mice, or whether the differences are somewhat consistent amongst different groups. It's possible that this is all accounted for with the various tests they do testing for normality of the data etc, but that goes beyond what I'm able to understand.

What doesn't inspire too much confidence in me is that the fibromyalgia findings that the authors cite seem to have turned out to be nothing burgers or at least have not resulted in anything tangible for patients. However, this groups newer LC grant seems to be centered around identifiying the actual antibodies, so we should find out relatively soon whether the findings hold any water, as I guess that this is the more important step.

 

@Jonathan Edwards you had significant reservations when reading the Fibromyalgia study by Goebel et al. Do you have similar thoughts here?

 

Summary thread on X / Twitter by one of the authors:

https://twitter.com/user/status/1796901736151392282


Interestingly, after separating the patients into groups based on a proteomics analysis of their blood, they found that the IgG of one group induced delayed symptoms:
https://twitter.com/user/status/1796901749728464934

 

I would contrast this paper with the sort of paper that one found in a journal of 1974 - where a single simple experiment was reported and in six cases out of ten became a classic piece of evidence that everyone learnt about. These days we see thirty seven graphs in colour about various things not directly related to the main experiment but not much detail on the finding of interest. In 99 out of 100 cases the papers are forgotten.

I am not aware of any disease in which transfer of human antibodies to mice has been critical in our understanding. Human immunoglobulin in mice will form immune complexes, fix complement, activate Fc receptors and generally cause a nuisance, whatever it is against.

I just don't see this sort of paper as a serious attempt to convey a simple, sound, scientific observation. It is more like what pops up when you try to read a newspaper - all sorts of flashing advertisements here there and everywhere making it impossible tor read what you wanted to look at. It is bad manners apart from anything.

We all used to do bad experiments in the 1970s but didn't bombard people with all this stuff and didn't expect anyone to take much notice. Now people spend millions of dollars on studies like this and nobody has any idea what to make of it.

 

The data on mechanical and heat sensitive in the mice seems to have the clearest effect (shown in figure 5 below), one that was present even if you ignore the subgrouping they used.

The effect on immobility and locomotor ability was only present in one of the three subgroups on day 1 (not on day 2, 3, 4 etc.). It is a bit unclear how they determined these subgroups: the selection seems to be quite complex and based on multiple cytokines and proteins: was this a post-hoc subdivision? [EDIT: probably not given that they pooled the IgG per subgroup]

 

Some of the figures such as figure 5 above note that they used 'BH adjustment' - so Benjamini-Hochberg adjustments. Unfortunately no exact p-values or effect sizes are given.

In this study they used a control group, so it's quite interesting that their antibodies did not have the same effect in mice as the antibodies of LC patients. For some reason the LC antibodies seem to be causing more nuisance.

 

“Another cool thing: our findings seem to have been independently validated by the @PutrinoLab, who recently showed similar preliminary data on this scientific conference (May 2024): youtube.com/watch?v=F3Sv9t… 10/11”

 

Having once been an 'animal model authority' I am pretty sceptical about findings like that. There are so many things that can affect your results. Unless the studies are fully blinded (which they rarely are) and repeated with the same spread of results it is anybody's guess what caused them.

 

Naive question: when transferring IgG like this is it certain that pure unbound IgG is all that's transferred? Can anything else be going along for the ride?

 

I would love to hear a fair critique of the methodology used in this paper by you, with this valuable information the methodology might actually end up being improved for the upcoming MECFS research this group plans to do. Exchanging information/knowledge to improve outcomes for pwME!

 

It is hard to wade through it all.
Two things I noted:

They pooled sera. That seems to me a disaster. If you pool 8 sera and give them to 8 mice then you may have information that at least 1 in 8 of the sera has something in it. If you took just two pure sera and gave them to 4 mice each then you would know if 2 out of 2 had something in it.

I also note that patient samples were sera and control samples plasma. Lack of standardisation at that level raises concerns.

Almost anything can be carried over with IgG. Remember that human IgG is not mouse IgG and is likely to be cleared pretty quickly by forming immune complexes with mouse proteins.

My main concern is that after 50 years I am not sure that anyone has ever convincingly transferred a human disease this way. Passive transfer was a trendy new technique in 1975 but I think we concluded it does not work. So this would not be using a well validated technique to prove a biological mediator. It would be using a technique that fell out of favour long ago.

 

The controls aren't really there.. would have been nice to see what IgG from healthy with an infection does.

Plus the pooling really lowers the power.

Kind of just suggests that elevated inflammation is not good for you. But maybe this is a good model for future work where they get enough IgG from individuals.

 

If it was a good model I think drug companies would have sorted out the methodology and been using it for screening drugs for decades. The kinetics and the immunochemistry don't really make any sense to me anyway.

 

The Iwasaki/Putrino group look to have done that, per recent presentation of unpublished data. (Doesn't look like significant difference between acute and Long Covid for balance/co-ord, grip strength and pain.)

Edit: as MelbME points out below there is a difference for pain (p < 0.0081).

 

That's cool, where is that from?

Oh I just saw the embedded link.

 

Screengrabs from the SolveME presentation 3 weeks ago. At 16:10 below.

https://www.youtube.com/watch?v=F3Sv9tfu_ww

 

The hot plate test shows a difference between Long COVID and acute COVID. Heat intolerance is a sign of dysautonomia. Perhaps anti-M3. Interesting to know if they also pooled, I'm guessing they probably did based on sample size.