Negative Association of [GWI] Symptomatology with Predicted Binding Affinity of Anthrax Vaccine Antigen to [HLA II molecules], 2025, James et al

[forestglip]

Negative Association of Gulf War Illness Symptomatology with Predicted Binding Affinity of Anthrax Vaccine Antigen to Human Leukocyte (HLA) Class II Molecules

Lisa M James, Apostolos P Georgopoulos

Background
Anthrax is a serious disease caused by Bacillus anthracis (B. anthracis) with a very high mortality when the spores of B. anthracis are inhaled (inhalational anthrax). Aerosolized B. anthracis spores can be used as a deadly bioweapon. Vaccination against anthrax is the only effective preventive measure and, hence, the anthrax vaccine was administered to United States (and other) troops during the 1990–91 Gulf War. However, the anthrax vaccine is not harmless, and the anthrax vaccination has been linked to the occurrence and severity of Gulf War Illness (GWI), a debilitating Chronic Multisymptom Illness (CMI). We hypothesized that this is partly due to the combination of two factors, namely (a) the cytotoxicity of the antigen (anthrax Protective Antigen, PA) contained in the vaccine, and (b) the Human Leukocyte Antigen (HLA) genotype of susceptible vaccinees, reducing their ability to make antibodies against the cytotoxic PA.

Method
Here, we tested this hypothesis by determining the association between severity of GWI symptoms in 458 GW veterans and the overall strength of the binding affinity of the PA epitopes to the specific six Human Leukocyte Antigen (HLA) Class II alleles carried by each individual (two of each of the HLA-II genes: DPB1, DQB1, DRB1), responsible for initiating the process of antibody production in otherwise immunocompetent individuals, estimated in silico.

Results
We found that the severity of GWI symptomatology was negatively and significantly correlated with the strength of the predicted binding affinity of PA peptides to HLA-II molecules (r=−0.356, p<0.001); the stronger the overall binding affinity, the weaker the symptoms. Since the binding of a peptide to an HLA-II molecule is the first and necessary step in initiating the production of antibodies, the findings above support our hypothesis that the severity of GWI symptomatology is partly due to a lack of HLA-II protection.

Conclusions
Reduced HLA protection against the toxic anthrax vaccine may underlie GWI.

Link | PDF (Vaccines) [Open Access]

 

[forestglip]

This is pretty interesting. They got the protein sequence of the anthrax vaccine, as well as the HLA genetic code of all the participants. They used a program called NetMHCpan where you input both and it predicts how well the peptide would bind to the MHC (MHC is the protein complex that HLA codes for). They used all consecutive 15 amino acid sequences from the full vaccine protein (750 sequences). They used a threshold score from the predictor, below which they considered it a "hit" (good binding), and above, no hit. Since each person has 6 versions of MHC class II, each person got 6 scores which were added together.

(Edit: to clarify, each person had 6 HLA sequences. Each HLA was tested against each of the 750 peptides. So each person's total score was the sum of hits for 6*750=4500 HLA-peptide pairs.)

There was a correlation where the more hits a person had (the more good peptide/MHC bindings the program predicted for them) the lower their GWI score (less disabled). (r = -0.356, p < 0.001)


This appears to be a replication of sorts of their previous study (1) which was on a smaller group of 42 veterans with GWI and it included 69 without GWI. There they found that getting the anthrax vaccine was associated with getting GWI. They also did something similar to the thread's study where they looked at HLA alleles that the people who got vaccinated but did not develop GWI had, but the ones who were vaccinated and did develop GWI did not have. These alleles the healthy people had were predicted by a computer program to bind very effectively to the vaccine peptides.

Other interesting bits:

Remarkably, the very same GWI symptomatology was observed in veterans tested one year post-deployment to Iraq (Operation Iraqi Freedom, OIF) and Afghanistan (Operation Enduring Freedom, OEF) wars [27,28] which were twenty years after the 1990–91 Gulf War. More specifically, 49.5% of these veterans met the criteria for mild-to-moderate CMI and 10.8% met the criteria for severe CMI [10]. Moreover, “Over 90% of Veterans with chronic pain met criteria for CMI. CMI was not completely accounted for either by posttraumatic stress disorder or by predeployment levels of physical symptoms. Veterans with symptoms consistent with CMI reported significantly worse physical health function than Veterans who did not report symptoms consistent with CMI” [27] (p. 59). These are alarming statistics. Presumably, soldiers deployed to OIF/OEF received the anthrax vaccine, which was mandatory for troops deployed to the Middle East. Various environmental exposures unique to the 1990–91 Gulf War (e.g., exposure to burning pits) that have been blamed for GWI [29] did not occur in the OIF/OEF operations and, therefore, cannot account for the CMI symptoms associated with these wars.

Previous research documented that the presence of certain Class II alleles, but no Class I alleles, discriminated between healthy Gulf War veterans and those with GWI such that healthy veterans possessed certain Class II alleles that were absent or significantly less frequent in those with GWI [6]. Subsequent research found that of the 69 HLA Class I alleles investigated, none were characterized by the combination of high-affinity binding to PA epitopes and high immunogenicity to anthrax vaccine antigen whereas several HLA Class II alleles were [ 2]. The unique effect of Class II HLA on GWI reported now in several studies suggests that protection against GWI or lack thereof, at least with regard to anthrax vaccination, sits squarely with antibody production, the very role of Class II HLA.

This refers to reference 2 below:

Crucial evidence for the presence of PA in the serum of veterans with GWI has been provided by [...] the addition of serum from GWI patients to neuroblastoma N2A cultures induced decreased neurite spreading and cell death, both of which were partly reversed by the addition of anti-PA antibodies [30].

1. Anthrax Vaccination, Gulf War Illness, and Human Leukocyte Antigen (HLA), 2024, James et al
2. Anthrax and Gulf War Illness (GWI): Evidence for the Presence of Harmful Anthrax Antigen PA63 In the Serum of Veterans with GWI, 2019, Tsilibary et al

 

[forestglip]

They just recently published another paper which is an opinion suggesting that poor HLA binding to SARS-CoV-2 allows it to persist and cause long COVID (1). Maybe also something similar happening with the COVID vaccine not binding well to HLA, since people report that the vaccine seems to cause ME/CFS on its own as well.

1. Human Leukocyte Antigen (HLA) at the Root of Persistent Antigens and Long COVID, 2025, Georgopoulos et al (thread)

 

[Hutan]

I haven't read this yet. It sounds interesting, especially the bit about soldiers effectively getting Gulf War Illness but not having been in the Gulf War and not having been exposed to sarin.

Both authors are from

The GWI and HLA Research Groups, Brain Sciences Center, Department of Veterans Affairs Health Care System, Minneapolis, MN 55417, USA;

It's good to see researchers from the US Dept of Veterans Affairs taking GWI seriously.

I wonder about the mechanism for antigens persisting though.

In an ideal situation, when the HLA-II genotype of the vaccinee provides molecules that can bind with high affinity to antigen epitopes, antibodies will be made and protection conferred.

On the other hand, if these molecules happen not to possess the requisite affinity to bind with the antigen epitopes, two consequences follow. First, antibodies are not made, or are produced in inadequate quantities, hence protection is inadequate; second, the vaccine antigen is not eliminated and may persist for a while, depending on the antigen and the
condition of the host.

This makes it seem that it is the lack of antibodies to an antigen that allows it to persist. I don't think we have seen a lack of antibodies to Covid-19 in post-Covid-19 ME/CFS.

 

[forestglip]

This makes it seem that it is the lack of antibodies to an antigen that allows it to persist. I don't think we have seen a lack of antibodies too Covid-19 in post-Covid-19 ME/CFS.

Good point. Do you know that this has been tested? They don't mention observed antibody levels in the long COVID paper.

Maybe it hinges on how fast antibodies start getting produced - maybe everyone gets antibodies by day 10, but if they didn't have them by day 5, the virus had a chance to find a reservoir to hide in?

 

[forestglip]

This one seems to support it, though I haven't thoroughly read it:

Antibody Response to SARS-CoV-2 is Associated with Long-term Clinical Outcome in Patients with COVID-19: a Longitudinal Study, 2021, García-Abellán et al

Long-COVID is associated with weak anti-SARS-CoV-2 antibody response, severity of illness, and female gender.

----

Edit: Another one. Not sure what to make of nucleocapsid antibody being higher in LC, but spike antibody being lower.

SARS‐CoV‐2 antibody levels and long COVID occurrence in blood donors, 2024, Avelino‐Silva et al

Anti-SARS-CoV-2 antibodies were time-dependently associated with long COVID; higher anti-nucleocapsid levels were associated with higher risk; and higher anti-spike levels were associated with lower risk of long COVID.

Maybe spike antibodies are the optimal tool to fight the virus, and people with LC don't produce as many. Virus levels get really high in LC, and because of that lots more nucleocapsid Abs are produced, but they aren't as useful.

 

[Hutan]

Sorry, I nodded off for a while. Back again.

I'd formed the impression that SARS-CoV-2 antibody levels weren't different in Long Covid, although interesting that the studies you found @forestglip suggest otherwise.

That chart that @forestglip posted, the relationship between GWI score and the sHm actually isn't that impressive. Apart from a couple of outliers, the values are all over the place.



GWI score

2.2. GWI Status and Severity

GWI symptom severity was assessed in the following 6 domains: fatigue, pain, neurological/cognitive/mood, respiratory, gastrointestinal, and dermatologic [3]. Only symptoms of at least moderate severity, that began during or after the Gulf War and lasted > 6 months counted towards a particular symptom domain. The overall GWI symp- tom severity score was the average of severity scores in the 6 symptom domains above.

The authors don't say how they selected their participants, just that they were US Gulf War veterans. It is interesting that there are only a handful of veterans with a GWI symptom score of zero and then a wide range of scores above that. It definitely is not a case of GWI or healthy. That's not necessarily bad.

sHm
First they worked out what HLA-II alleles each veteran had (each person has 6). They found 93 different alleles.

Then they worked out the binding efficiency of each HLA-II allele to the antigen from the vaccine. (They tested 750 different chopped up bits of the antigen against each HLA-II allele, which sounds mind-boggling).

That then produced a binding efficacy score, the sHm.


But yeah. Look at that chart. I can't really see much of a relationship there. It's certainly not the whole answer.

 

[Hutan]

There was a correlation where the more hits a person had (the more good peptide/MHC bindings the program predicted for them) the lower their GWI score (less disabled). (r = -0.356, p < 0.001)

 

[forestglip]

They tested 750 different chopped up bits of the antigen against each HLA-II allele, which sounds mind-boggling

They didn't actually test real peptides in the lab. They used the NetMHCpan neural network program. You give it the DNA code of the HLA and the amino acid letters of the peptide, and it predicts how well they would bind.

That chart that @forestglip posted, the relationship between GWI score and the sHm actually isn't that impressive. Apart from a couple of outliers, the values are all over the place.

But yeah. Look at that chart. I can't really see much of a relationship there. It's certainly not the whole answer.

Well we have stats to actually put numbers on the relationship. It's clearly not a one to one relationship visually from the chart, but with a p value less than 0.001, it's very likely the observed association isn't due to chance. And there is a moderate correlation of -0.356. It's not a super high correlation, but I don't think they suggest that anthrax vaccine is definitely the whole story for everyone with GWI.

I only just learned about this NetMHCpan program so I don't know how accurate these binding scores are, but at the very least it seems there's an association between HLA in some way and GWI score.

 

[Eddie]

If you removed the one outlier with a very high GWI score and the two individuals with more than 150 hits, I 'm guessing the correlation would only be very weak. There could still be an association, but it would have been more convincing if the trend wasn't driven by a handful of individuals.

 

[Amw66]

There were multiple vaccinations, with more than one vaccine administered concurrently .

There are also.different adjuvants

It may be more enlightening if it were possible to mimic the vaccination protocols .

 

[forestglip]

If you removed the one outlier with a very high GWI score and the two individuals with more than 150 hits, I 'm guessing the correlation would only be very weak. There could still be an association, but it would have been more convincing if the trend wasn't driven by a handful of individuals.

Sounded fun to try to test. I digitized the image of the plot with WebPlotDigitizer. They're not perfectly on center for each dot, but I'm pretty sure I got every dot's approximate location. The paper says there are 98 dots, and I was sure about 97 of them, and pretty sure about the last. Points I placed at each of the dot locations:


So first I plotted and did a Pearson test on the full dataset, like what the authors did:

They said r = -0.356, I got r = -0.360. Not perfect, but I'd call that pretty much the same. And I got p=0.00027.

I removed the three points you suggested and ran the stats again:

r = -0.294, p = 0.0038

That's still a pretty low p value. And removing these outliers is skewing in the opposite direction by not factoring them in at all. The correlation didn't decrease that much, and is just below the Pearson rule of thumb cutoff of 0.3 for "moderate" relationship.

 

[EndME]

Knowing nothing about GWI I had been under the impression that there were some primilarly findings heading in the direction of genetics and "something to do with nerves". So I'm a bit curious where their hypothesis for the anthrax vaccine came from. If it was a true association shouldn't it also show up amongst UK veterans (seemingly those were also given an anthrax vaccine)?

I wander how they recruited for their study and whether that introduced bias (for instance if the study would have recruited used something along the lines of "Study of Gulf War illness and anthrax vaccine"). Since the larger part of vaccinated veterans didn't develop GWI wouldn't it also have made sense to look at participants that didn't receive a vaccine but are classified to having GWI to see if there are stronger signals in the genes? After all even if a true signal it might have nothing to do with the vaccine...

Given the poor quality of research we've seen in LC and ME/CFS I wouldn't be surpised if there's some substantial bias in the samples even if the definition of GWI is in line with the IOM.

 

[forestglip]

So I'm a bit curious where their hypothesis for the anthrax vaccine came from.

They've got a few papers on this. I might see if one of the other ones shows where it started later.

If it was a true association shouldn't it also show up amongst UK veterans (seemingly those were also given an anthrax vaccine)?

Would be interesting to check. I forgot I was also planning to see if there are any safety studies on this vaccine that might show something like increased fatigue after a year.

I wander how they recruited for their study and whether that introduced bias (for instance if the study would have recruited used something along the lines of "Study of Gulf War illness and anthrax vaccine").

I'm not sure how this would introduce bias. The independent variable is genetics, so unless these people intimately know their HLA DNA, I don't see how knowing it's about anthrax vaccine would influence the sample.

Since the larger part of vaccinated veterans didn't develop GWI wouldn't it also have made sense to look at participants that didn't receive a vaccine but are classified to having GWI to see if there are stronger signals in the genes? After all even if a true signal it might have nothing to do with the vaccine...

Good idea. Maybe they'll eventually do that.

I wouldn't be surpised if there's some substantial bias in the samples even if the definition of GWI is in line with the IOM.

Again, I'm not really sure what exactly the bias might look like, maybe you can clarify?

 

[EndME]

Again, I'm not really sure what exactly the bias might look like, maybe you can clarify?

From what I've understood from their earlier papers their whole hypothesis is based on a sample where having had this vaccine correlated with GWI status and severity. The question is whether that is a true relationship or just a correlation within their sample set. Other studies have found a relationship between GWI status and severity and genetics, whilst other studies have picked up all sorts of other correlations, never leading to anything causal. So GWI status and genetics can't be assumed to be independent. If you recruit people along the lines of "Do you have GWI and did you receive an anthrax vaccination?" it's plausible that you end up with more severly effected people in the anthrax group. That could explain results of the first study and their hypothesis (where they looked at GWI status and HLA-II alleles amongst anthrax vaccinated veterans). At the same time more severly effected people might be more likely to have certain genetic HLA-II alleles. This could be the correlation seen in this study since I haven't seen anbody comment on whether NetMHCpan bears any relevance. Perhaps it's even possible that people with certain HLA-II alleles are more likely to report something else that makes them severe (comorbidities etc aren't controlled for, perhaps some of them have MS or are more likely to have certain autoimmune conditions which are linked to certain HLA-II alleles) and that this is what NetMHCpan also somehow partially captures.

Again, I'm not really sure what exactly the bias might look like, maybe you can clarify?

Nobody knows but I would think there's a tremendous amount of room for bias in a genetic study with a tiny sample size. There's lots of studies finding a link between Long-Covid status and genetics but it is unclear whether any of those have anything to do with Long-Covid.

A BPS researcher might argue that these genes are simply linked to being "depressed" and NetMHCpan simply finds an associated relationship to depression (coincidentally or not). Perhaps these are simply genes linked to being more likely to answering a phone call about a study (and coincidentally or not linked to NetMHCpan's analysis)...

 

[forestglip]

If you recruit people along the lines of "Do you have GWI and did you receive an anthrax vaccination?" it's plausible that you end up with more severly effected people in the anthrax group. That could explain results of the first study

Yeah the first study could easily be biased.

At the same time more severly effected people might be more likely to have certain genetic HLA-II alleles.

Perhaps it's even possible that people with certain HLA-II alleles are more likely to report something else that makes them severe (comorbidities etc aren't controlled for, perhaps some of them have MS or are more likely to have certain autoimmune conditions which are linked to certain HLA-II alleles) and that this is what NetMHCpan also somehow partially captures.

A BPS research might argue that these genes are simply linked to being "depressed" and NetMHCpan simply finds an associated relationship (coincidentally or not). Perhaps these are simply genes linked to being more likely to answering a phone call about a study (and coincidentally or not linked to NetMHCpan's analysis)...

So I'm mainly intrigued because I think we had seen a couple studies already finding associations between HLA and long COVID. Though my memory is foggy, I'll have to look.

I'm not thinking too much about whether NetMHCpan shows what is says it does. It'd be interesting if their hypothesis is right, but even if the program, behind the scenes, is a four year old saying which DNA letters they like better, it's another association of HLA and one of these similar chronic illnesses. Even if the HLA gene causes depression which resembles GWI, I'd want to know.

Though I do have to see if I'm remembering right about the other HLA studies.

 

[forestglip]

Nobody know but I would think there's a tremendous amount of room for bias in a genetic study with a tiny sample size.

I don't think this is a tiny sample size. I think genetics studies generally need enormous samples when they are testing massive numbers of genes or the whole genome, to have enough power to detect effects after multiple test correction, and they have to correct for thousands or millions of tests.

Here it's just one test, so I think this is actually a fairly large sample at 458 people.

 

[Jonathan Edwards]

This seems pretty odd.

If anthrax protein in a vaccine is toxic it is likely to be toxic within hours - maybe over 24hrs if it seeps out slowly.

Making an antibody response to an antigen takes about 10-20 days.
So it is very unlikely to affect toxicity of a bacterial antigen.

I would have expected if anything for people to blame the immune response to the antigen for GWI. They are suggesting the opposite here.

The way the abstract is written does not induce confidence in their immunological knowledge. I have never heard of anyone talk of HLA-II before. HLA antigens come as A,B,C, or D (also E,G). All these fall into the MHC complex (which is much larger and contains things like TNF genes). HLA-A,B and C are in the MHC-I segment and D is in the MHC-II segment.

They also say that peptide binding to class II is the first step in antibody production. It isn't by a long way. The whole antigen has to be taken up by a B cell that recognises it first, in order to be processed and presented to T cells. B cells can respond, survive and proliferate after recognising whole antigen without needing T cells. Moreover, in the time window of toxin action almost all antibody effects will be from IgM antibodies pre-existing the vaccine injection.

I wouldn't have let that abstract through peer review.

 

[Jonathan Edwards]

Here it's just one test, so I think this is actually a fairly large sample at 458 people.

Yes, 458 sounds fine. This isn't really a genetic study. It is a study of functions of proteins that happen to have a limited and well documented polymorphism.

 

[EndME]

So I'm mainly intrigued because I think we had seen a couple studies already finding associations between HLA and long COVID. Though my memory is foggy, I'll have to look.

You will find any type of association you want to find with Long-Covid because the definition is so vague that any results depend on the sample you wish to choose.

Even if the HLA gene causes depression which resembles GWI, I'd want to know.

I would agree! But without rigorously methodology you have no idea where the noise might be coming from...If HLA-II alleles are linked to sex (which they aren't as @Jonathan Edwards comments below and which clearly shows how little I know about any of these topics), then they could be linked to people being more responsive to answering a phone call advertising a study (if females are more likely to stay at home even if those reasons are purely societal) or because they are more likely to be in hospitals where they see study advertisements (because they have comorbidities linked to certain genes). Now the one example (sex bias) isn't relevant to this study, but who knows what might be...