HLA and pathogens in myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) and other post-infection conditions, 2025, Georgopoulos et al

[Wyva]

Apostolos P. Georgopoulos, Lisa M. James & Philip K. Peterson

Abstract​

Viral infections have been widely implicated in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) pathogenesis. Recent evidence has also identified certain Human Leukocyte Antigen (HLA) alleles that are significantly associated with ME/CFS risk/protection. Here we tested the hypothesis that ME/CFS risk or protection conferred from those HLA alleles is associated with binding affinity to antigens of HHV viruses, a critical step in initiating the adaptive immune system response to foreign antigens.

Specifically, we determined in silico the predicted binding affinity of two susceptibility alleles (C*07:04, DQB1*03:03) and two protective alleles (B*08:01, DPB1*02:01) to > 10,000 antigens of the 9 Human Herpes Viruses (HHV1, HHV2, HHV3, HHV4, HHV5, HHV6A, HHV6B, HHV7, HHV8) which have been implicated in the etiology of ME/CFS. We found that the binding affinity of all HHV antigens to the susceptibility alleles was significantly weaker than the binding affinity to the protective alleles (P < 0.001). In fact, none of the HHV antigens showed strong binding to the susceptibility alleles, in contrast to the strong bindings showed by the protective alleles. These findings are in keeping with the hypothesis that the effect of a putative HHV insult in contributing to ME/CFS is modulated by the host’s HLA immunogenetic makeup.

We speculate that strong HLA-antigen binding likely protects against ME/CFS via elimination of virus antigens; conversely, weak HLA-antigen binding may permit persistence of foreign antigens, contributing to ME/CFS and other chronic conditions. Finally, with respect to the latter, we determined the binding affinities to the 4 HLA alleles above to pathogens causing two chronic diseases with very similar symptomatology to ME/CFS, namely Long COVID and post-treatment Lyme disease syndrome (PTLDS).

We found that the 2 ME/CFS susceptibility HLA alleles above had very weak binding with SARS-CoV-2 virus glycoprotein (involved in Long COVID) and 5 proteins of Borrelia burgdorferi (involved in PTLDS), in contrast to the ME/CFS protective alleles that showed strong bindings. These findings support the hypothesis that ME/CFS, long COVID and PTLDS are caused by persistent pathogenic antigens that could not be eliminated due to inadequate protection by the patient’s HLA makeup.

Open access: https://www.nature.com/articles/s41598-025-21230-z

 

[V.R.T.]

Apostolos P. Georgopoulos, Lisa M. James & Philip K. Peterson

Abstract​

Viral infections have been widely implicated in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) pathogenesis. Recent evidence has also identified certain Human Leukocyte Antigen (HLA) alleles that are significantly associated with ME/CFS risk/protection. Here we tested the hypothesis that ME/CFS risk or protection conferred from those HLA alleles is associated with binding affinity to antigens of HHV viruses, a critical step in initiating the adaptive immune system response to foreign antigens.

Specifically, we determined in silico the predicted binding affinity of two susceptibility alleles (C*07:04, DQB1*03:03) and two protective alleles (B*08:01, DPB1*02:01) to > 10,000 antigens of the 9 Human Herpes Viruses (HHV1, HHV2, HHV3, HHV4, HHV5, HHV6A, HHV6B, HHV7, HHV8) which have been implicated in the etiology of ME/CFS. We found that the binding affinity of all HHV antigens to the susceptibility alleles was significantly weaker than the binding affinity to the protective alleles (P < 0.001). In fact, none of the HHV antigens showed strong binding to the susceptibility alleles, in contrast to the strong bindings showed by the protective alleles. These findings are in keeping with the hypothesis that the effect of a putative HHV insult in contributing to ME/CFS is modulated by the host’s HLA immunogenetic makeup.

We speculate that strong HLA-antigen binding likely protects against ME/CFS via elimination of virus antigens; conversely, weak HLA-antigen binding may permit persistence of foreign antigens, contributing to ME/CFS and other chronic conditions. Finally, with respect to the latter, we determined the binding affinities to the 4 HLA alleles above to pathogens causing two chronic diseases with very similar symptomatology to ME/CFS, namely Long COVID and post-treatment Lyme disease syndrome (PTLDS).

We found that the 2 ME/CFS susceptibility HLA alleles above had very weak binding with SARS-CoV-2 virus glycoprotein (involved in Long COVID) and 5 proteins of Borrelia burgdorferi (involved in PTLDS), in contrast to the ME/CFS protective alleles that showed strong bindings. These findings support the hypothesis that ME/CFS, long COVID and PTLDS are caused by persistent pathogenic antigens that could not be eliminated due to inadequate protection by the patient’s HLA makeup.

Open access: https://www.nature.com/articles/s41598-025-21230-z

HLA-C and DQB1!

Does the conclusion they come to about persistent herpesvirus line up with the data? It doesn't look like they determined the binding affinity to other types of antigen.

Edit: my mistake, missed the first part of that last paragraph.

 

[forestglip]

HLA-C and DQB1!

They specifically looked at these and the two other protective alleles because of the previous study that identified them.

 

[V.R.T.]

My brain fog is bad right now but couldn't these results suggest weak binding with all antigens rather than pathogen specific ones? As in immune dysfunction not viral persistance as the driver.

 

[V.R.T.]

They specifically looked at these and the two other protective alleles because of the previous study that identified them.

Yes, I misread the abstract somewhat!

 

[ME/CFS Science Blog]

Here's what I think the study did: A previous Norwegian study found a link between HLA-variants and ME/CFS. Two were risk factors (more common in ME/CFS) and two were protective (more common in controls).

This study checked how well these 4 variants bind to human-herpes viruses (HHV). Turns out that the ME/CFS risk variants bind less well to HHV than the variants that were protective. In other words, the variants that were more common in ME/CFS patients were less affective at targetting these viral particles. They also tested and got the same results for peptides of the SARS-CoV-2 spike glycoprotein (the virus that causes COVID-19) and B. burgdorferi (the bacterium that causes Lyme Disease).

The only problem: the DecodeME study could not confirm the Norwegian results. They plan to redo the HLA-analysis, so hopefully that might change the results for this region. But if not, I'm afraid that would also make the findings of this study less interesting. It would suggest that these HLA-variants aren't really associated with ME/CFS after all.

 

[Jonathan Edwards]

I see the problem here being that the association with HLA C 0704 and HLA-DQB1 0303 didn't show up clearly on DecodeME so it looks as if it is very weak even if bona fide.

That tends to lead to the conclusion that differential binding of thousands of herpes antigens isn't actually important in ME/CFS.

 

[forestglip]

Finally, we evaluated the association between the log-transformed odds ratio of ME/CFS12 and the log-transformed mean PBBA of the 9 HHVs (Table 4) and found a strong correspondence (Pearson = 0.956, P = 0.044, N = 4).

For the correlation, I think the sample size of 4 points might be too small to give a reliable p-value.

When I test pearson correlation in R with those four points, I get the same result for correlation (0.956) and p-value (.044) as the study got. But when I test the correlation using a permutation method instead, the p-value is much larger at p=~0.33. I think this is more reliable for the tiny sample size.

 

[Utsikt]

If we assume that these findings actually are relevant for ME/CFS, could the weak binding to HHV, Covid and BB lead to relatively more binding to something else? Like random junk?

 

[V.R.T.]

If we assume that these findings actually are relevant for ME/CFS, could the weak binding to HHV, Covid and BB lead to relatively more binding to something else? Like random junk?

This is what i was trying to get at earlier

 

[forestglip]

If we assume that these findings actually are relevant for ME/CFS, could the weak binding to HHV, Covid and BB lead to relatively more binding to something else? Like random junk?

Maybe, but I think it's a bit of a leap from this data.

The simplest explanation to me, if the results turned out to be reliable, is that in people with ME/CFS, their HLA/MHC binds less strongly to certain pathogens, and thus less of an adaptive immune response can be raised against the pathogens.

If more severe infections increase risk of developing ME/CFS (which I don't know for a fact, but seems intuitively plausible), then I can imagine that a worse adaptive response (from weak HLA binding) might lead to a more severe infection which might increase risk.

 

[Jonathan Edwards]

Another worry is that the putative protective role of B8 in the Norwegian data was linked to having the A1, B8, Dr3 haplotype if I remember rightly. That suggests that it isn't necessarily anything to do with B8 itself. And although Chris thought he might have had the same result it turned out to be an artefact.

 

[V.R.T.]

I see the problem here being that the association with HLA C 0704 and HLA-DQB1 0303 didn't show up clearly on DecodeME so it looks as if it is very weak even if bona fide.

That tends to lead to the conclusion that differential binding of thousands of herpes antigens isn't actually important in ME/CFS.

Don't we have to wait for the HLA analysis before we can conclude that they didn't show up? I thought the whole point of that is that the normal GWAS method can miss a lot about HLA.

 

[Jonathan Edwards]

Don't we have to wait for the HLA analysis before we can conclude that they didn't show up? I thought the whole point of that is that the normal GWAS method can miss a lot about HLA.

My reading of what I have been told about the data is that they could have picked out individual risk and protective alleles with precision but in the end the analysis showed that initial analysis was being skewed by an artefact. What seems pretty clear is that we are not seeing the sort of linkage that we see in diseases like RA and ank spond. Chris was talking of differences in rates of alleles in populations of something like 23 vs 20 percent. To me it is just not credible that this indicates that bidning of specific viral peptides is critical. It clearly isn't.

Everyone looking at HLA immediately thinks in terms of screening peptide binding. I am sceptical that this is ever very useful. In RA it may be that binding of citrulline is crucial but there are millions of peptides that can be citrullinated. In any spond the impact on infection resistance cuts across diverse organisms with many different antigens. I suspect that more time has been wasted screening for peptide binding profiles than anything else in immunology since 1990.

And another thins is that rather few T cells are DQ restricted, so DR is much more likely to be relevant.

 

[forestglip]

Another worry is that the putative protective role of B8 in the Norwegian data was linked to having the A1, B8, Dr3 haplotype if I remember rightly. That suggests that it isn't necessarily anything to do with B8 itself. And although Chris thought he might have had the same result it turned out to be an artefact.

Here's the relevant part of that study:

Human Leukocyte Antigen alleles associated with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS), 2020 (thread)

There were two alleles with a negative association with ME/CFS, suggesting a potential protection, namely B*08:01 (OR = 0.7 [95% CI 0.6–0.9], pnc = 0.01, pc = n.s.) and DPB1*02:01 (OR = 0.7 [95% CI 0.6–0.9], pnc = 0.02, pc = n.s.) (Table 2). These alleles were not in LD (D’ = −0.29), indicating that the associations are independent.

The most frequent B*08:01 haplotype in Norway is the highly conserved so-called autoimmune and ancestral AH8.1 haplotype31 (C*07:01-B*08:01-DRB1*03:01-DQB1*02:01). This haplotype had reduced estimated frequency in the patient group compared to the control group (8.2% vs. 10.3%, OR = 0.8, pnc = 0.06), albeit not significantly.

HLA-B*08:01 showed reduced frequency in ME/CFS compared to controls in our material. This allele most often occur on the haplotype C*07:01-B*08:01-DRB1*03:01-DQB1*02:01, which was also less prevalent among ME/CFS patients in our material.

This ancestral haplotype, AH8.1, is a risk factor for a wide variety of AID, including myasthenia gravis, systemic lupus erythematosus and coeliac disease31, but protective against rheumatoid arthritis34,35. In the existing literature on HLA and CFS, HLA-DRB1 is the locus most frequently studied.

In four out of five studies, the frequency of DR3/DRB1*03 was lower in the patient group23,24,25,28, while in the fifth study the frequency was similar in both groups22. Hence, this haplotype seems truly less prevalent among ME/CFS patients.

So B*08:01 by itself was significantly protective (only when not corrected for multiple tests, p=.01), while the AH8.1 haplotype did not quite pass the significance threshold (uncorrected p=.06). But close enough that I think it's still a concern that if the B*08:01 signal is real, the relevant allele for ME/CFS might be something else on the haplotype.

 

[V.R.T.]

they could have picked out individual risk and protective alleles with precision but in the end the analysis showed that initial analysis was being skewed by an artefact.

Could you possibly clarify what you mean by this for laypeople like myself?

And another thins is that rather few T cells are DQ restricted, so DR is much more likely to be relevan

Is DR the HLA that DQ is usually linked to but wasnt in DecodeME?

 

[Simon M]

My memory is terrible, but I I am not sure the Norwegian study had found HLA alleles that reached statistical significance (merely that they were promising).

I wonder when the DecodeME HLA analysis will be done?

 

[forestglip]

My memory is terrible, but I I am not sure the Norwegian study had found HLA alleles that reached statistical significance (merely that they were promising).

I think the risk alleles were significant after multiple test correction, while the protective alleles were only significant without correction.

Risk alleles (pc: corrected p, pnc: not corrected p):

When comparing individual allele frequencies between patients and controls, four HLA risk alleles emerged (Table 2):

C*07:04 (OR = 2.1 [95% CI 1.4–3.1], pnc = 0.0001, pc = 0.001),
B*57:01 (OR = 1.6 [95% CI 1.2–2.3], pnc = 0.004, pc < 0.05),
DQB1*03:03 (OR = 1.5 [95% CI 1.1–2.0], pnc = 0.005, pc < 0.05) and
B*44:02 (OR = 1.3 [95% CI 1.0–1.6], pnc = 0.03, pc = n.s.).

Quote about the protective alleles posted above.

 

[forestglip]

They say everything to do the study is freely accessible, so it might be something to play around with.

Data availability

All information was obtained from freely accessible websites and, as such, is publicly and freely available. The antigen protein sequences obtained from Uniprot [ref.42: https://www.uniprot.org/uniprotkb] are provided in Table S1. The binding affinities of each 9-mer (HLA-I) and 15-mer (HLA-II) pathogen protein sequence to HLA was exhaustively tested using IEDB [ref.43: http://tools.iedb.org/mhci/]. The n-mer peptide sequence with the predicted best binding affinity for each HLA antigen is provided in Table S2.

 

[V.R.T.]

Could you possibly clarify what you mean by this for laypeople like myself?

So having reread this I think JE is talking about the inital Decode HLA analysis. And that it could well have picked out those connections but whatever it did pick up was discovered to be a computational hallucination so we were left with the DQ link.

And I checked and DR is the HLA protein closely associated with DQ.

Could the separate HLA analysis clear this up? Could we still find a DR link for our lonely DQ?

Did the other studies that previously found DQ find DR?