Preprint Role of the complement system in Long COVID, 2024, Farztdinov, Scheibenbogen et al.

[SNT Gatchaman]

Role of the complement system in Long COVID
Vadim Farztdinov; Boris Zuehlke; Franziska Sotzny; Fridolin Steinbeis; Martina Seifert; Claudia Kedor; Kirsten Wittke; Pinkus Tober-Lau; Kathrin Textoris-Taube; Daniela Ludwig; Clemens Dierks; Dominik Bierbaum; Leif Erik Sander; Leif Gunnar Hanitsch; Martin Witzenrath; Florian Kurth; Michael Muelleder; Carmen Scheibenbogen; Markus Ralser

Long COVID, or Post-Acute COVID Syndrome (PACS), may develop following SARS-CoV-2 infection, posing a substantial burden to society. Recently, PACS has been linked to a persistent activation of the complement system (CS), offering hope for both a diagnostic tool and targeted therapy.

However, our findings indicate that, after adjusting proteomics data for age, body mass index and sex imbalances, the evidence of complement system activation disappears. Furthermore, proteomic analysis of two orthogonal cohorts-one addressing PACS following severe acute phase and another after a mild acute phase-fails to support the notion of persistent CS activation. Instead, we identify a proteomic signature indicative of either ongoing infections or sustained immune activation similar to that observed in acute COVID-19, particularly within the mild-PACS cohort.


Link | PDF (Preprint: MedRxiv) [Open Access]

 

[SNT Gatchaman]

Refutes Persistent complement dysregulation with signs of thromboinflammation in active Long Covid (2024, Science)

Recently, a study by Cervia-Hasler et al. thus generated great interest as it linked the disease to a persistent activation of the complement system (CS), and therefore of the innate immune system, providing a causal explanation for PACS. The results raised hopes for a diagnostic tool and a targeted therapy.

Our curiosity was stirred by a result shown in Figure 8b, demonstrating that in their cohort, age and body mass index (BMI) alone predicted PACS with an area under the Receiver Operating Characteristic (ROC) curve of almost 0.8. This result stands in contrast with other studies in the field. While these agree that age and BMI can be risk factors of PACS, age and BMI alone are deemed insufficient to predict which individuals would develop PACS.

In seeking to explain why age and BMI are such strong predictors of PACS in this study, we noticed that the cohort was substantially imbalanced for age and BMI. The non-PACS control group predominantly consisted of young individuals with a median age of 36 years and a mean BMI of 25, while individuals in the PACS group were considerably older (median age of 58 years) and had a mean BMI of 28.

We were wondering whether this imbalance might have also affected other results.

 

[SNT Gatchaman]

To mitigate the age and BMI imbalances, we used a balanced factorial design strategy by splitting all patients in disease groups by age and sex. Because the cohort was too small to balance all parameters, we excluded those individuals with vastly different age and/or BMI. While this strategy reduced the cohort, it resulted in a reasonable balancing by age and BMI in subgroups. The final set consisted of 85 individuals (56 controls and + 29 PACS), with approximately the same ratio of PACS to non-PACS as the initial data set.

Comparing the proteomes between PACS and non-PACS in this balanced cohort, none of the complement components was significantly changed in PACS, using the same significance level set by Cervia-Hasler et al.

In contrast, this analysis revealed a different set of plasma proteins that emerged as discriminators of PACS. Apart from immunoglobulins, BCHE was significantly higher in males above 48 years of age, Paraoxonase 3 (PON3) was increased in PACS patients and Syntax Binding Protein 5 (STXBP5) was decreased in PACS patients.

 

[SNT Gatchaman]

subset having developed the most debilitating form of PACS, Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS). These patients are younger and predominantly female. We hence chose a third cohort which focuses on PACS with fatigue and exertional intolerance (“Charité post mild acute COVID19” cohort). This cohort included 53 PACS patients with persisting moderate to severe fatigue with 25 of them fulfilling CCC diagnostic criteria for ME/CFS.

A total of 27 non-PACS, Post Covid Healthy Controls (PCHC) were recruited during the same time. Serum proteome of participants was determined 5 to 19 months following acute COVID-19.

Statistical analysis of the Charité post mild acute COVID19 cohort (PACS (with ME/CFS and without) vs.– Post Covid Healthy Controls (PCHC)) revealed 27 differentially abundant proteins (at significance level alpha = 0.05 (fdr <= 0.33) and the estimated fold-change threshold (1.1, Figure 1F), of which 7 (CA1, SAA1, APOB, ITIH3, IGHV1-2, LRG1, CFI) are elevated. In this cohort, the only complement protein significantly upregulated, CFI, was correlated with BMI (r ~ 0.56).

Most of the 27 proteins that are significantly regulated by PACS were previously identified in COVID-19. Increased levels of SAA1 (Serum Amyloid A1), ITIH3 (Inter-Alpha-Trypsin Inhibitor Heavy Chain H3), CFI (Complement Factor I), LRG1 (Leucine-Rich Alpha-2-Glycoprotein 1) and decreased levels of Lumican (LUM), Fibronectin 1 (FN1), Multimerin-1 (MMRN1), Cartilage Oligomeric Matrix Protein (COMP), Transgelin 2 (TAGLN2), and Profilin 1 (PFN1) are consistent with protein regulation in severe acute patients.

The proteins mirror the response in inflammation, acute phase, coagulation, extracellular matrix as well as activation and degradation of platelets. The proteome in these individuals could therefore indicate a persistent infection or persistent immune activation.

 

[SNT Gatchaman]

GeneCards —

PON3
BCHE
STXBP5

Not sure about PON3 but PON1 was implicated in GWI: Evaluation of a Gene–Environment Interaction of PON1 and Low-Level Nerve Agent Exposure with Gulf War Illness: A Prevalence Case–Control Study Drawn from the U.S. Military Health Survey’s National Population Sample (2022, Environmental Health Perspectives).

Just looking at STXBP5, it has been associated with vWF, endothelial cells and platelets. Eg from Syntaxin-binding protein STXBP5 inhibits endothelial exocytosis and promotes platelet secretion (2014, The Journal of Clinical Investigation) —

The prominent relationship between human plasma vWF alterations and genetic variation of STXBP5 prompted us to investigate the potential role of STXBP5 in endothelial exocytosis.

STXBP5 was first discovered as a protein interacting with syntaxin 1A (STX1) in neuron and named tomosyn

STXBP5 inhibits neuron release of neurotransmitters and endocrine cell secretion of insulin or other vesicles

Here we showed that mammalian ECs express STXBP5. In vitro, STXBP5 interacted with the endothelial exocytic machinery and was a potent regulator of endothelial exocytosis.

We found that STXBP5 is expressed in human endothelial cells and colocalizes with and interacts with syntaxin. In human endothelial cells reduction of STXBP5 increased exocytosis of vWF and P-selectin. Mice lacking Stxbp5 had higher levels of vWF in the plasma, increased P-selectin translocation, and more platelet-endothelial interactions, which suggests that STXBP5 inhibits endothelial exocytosis.

Spoiler: Increased vWF and p-selectin have been shown in ME and LC (as well as the platelet-endothelial interactions). Eg recently

Blood–brain barrier disruption and sustained systemic inflammation in individuals with long COVID-associated cognitive impairment (2024, Nature Neuroscience)

Herpesvirus Infection as a Systemic Pathological Axis in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (2024, Preprint)

Neutrophil degranulation, endothelial and metabolic dysfunction in unvaccinated long COVID patients (2024, European Journal of Clinical Investigation)

 

[Hutan]

Thanks for posting @SNT Gatchaman, looks interesting, and a lesson in the need to match cohorts.

 

[Hutan]

Great that the investigators didn't stop at investigating the proteomics in the Zurich cohort after reassessing with the demographics properly taken into account, but also looked at two cohorts of their own: people post-severe Covid and people post-mild Covid.

Regarding that last cohort, the one likely to be most relevant to ME/CFS, the post-mild Covid:

This cohort included 53 PACS patients with persisting moderate to severe fatigue with 25 of them fulfilling CCC diagnostic criteria for ME/CFS (16). A total of 27 non-PACS, Post Covid Healthy Controls (PCHC) were recruited during the same time. Serum proteome of participants was determined 5 to 19 months following acute COVID- 19.

This study is more evidence that CRP isn't relevant in PACS.

 

[Hutan]

Has anyone looked at Supplementary Datafile 5? I don't know what the unit is for the proteins - all of the values seem to be in a very tight range, both between individuals and even between different proteins (ie approx 18 to 21). I assume the figures are the transformed outputs, log adjusted and probably also adjusted by the model that takes age and BMI into account?

There's also stuff about imputed values - from different visits? - in the methodology section.

I'm not up to working it all out, but if someone is across it, it could be interesting to hear about the process and understand it a bit more

 

[EndME]

Thanks for posting @SNT Gatchaman!

I didn't think highly of the study by Boyman et al to begin with, so I'm just wandering if Scheibenbogen et al also looked at analysing/comparing the results of the Cardiff study as well or whether those match their data of their own cohorts better (it seems they might have also had some BMI problems in that study which might have affected the results).

 

[Hutan]

As I said upthread, I think the post-mild Covid PACS fatigue+ cohort with some of the participants meeting ME/CFS criteria is the cohort of most interest to us.

While I love the overall approach of this paper, (i.e. attempting to account for the noise created by age and sex and BMI), I do think that they could have presented their results more clearly. I also wish that their post-mild Covid PACS cohort had all been compliant with the CCC (or another criteria with PEM). I hope the authors are as excited as I am by their results from the post-mild Covid PACS cohort and are continuing to work on getting some replicable proteomics for well characterised post-Covid ME/CFS.

Here's the volcano chart for the post-mild Covid cohort vs healthy controls, increased in size so the proteins can be read:

I don't know why the Supplementary Data tables don't give us the individual data with the same x axis unit.

Listing out the proteins of interest for the search engine:

Increased:
CFI, LRG1, APOB, IT1H3, CA1, SAA1, IGHV1-2

Decreased:
BTD, LTF, PON1, COMP, IGKVA-1, IGKV1-17, IGHV5-51, GP1BA, IGLV8-61, CFD, MMRN1, PTEN, PROZ, IGHV3-7, RBP4, IGLV1-51, LUM, FN1, PFN1, TAGLN2

Paraoxonase 3 (PON3) was increased in PACS patients

That's a clear illustration of how important it is to use a well characterised disease sample. Rather than PON3 being increased, as found in the initial cohort with its mixture of post-Covid consequences, it appears that PON1 was decreased in PACS-ME/CFS patients.

Looking for significant proteins is complicated in that it's not just a matter of proteins being elevated or decreased in the plasma. For one thing, it could be the levels of the protein in a particular tissue in the body that makes a difference, and the levels in plasma may or may not reflect that, or might even show the opposite quantity. As @SNT Gatchaman notes, PON1 is part of the GWI story. With the GWI work, it was found that different versions of the PON1 gene made people more or less vulnerable to neurotoxic sarin gas exposure, because one version is less good at detoxifying sarin. So, the researchers wouldn't necessarily find lower levels of the PON1 protein as a whole in the GWI patient's blood. What I think they found was a correlation between [the number of versions of the PON1 gene that was less good at detoxifying sarin (0,1,2) in the person's DNA] and [likelihood of developing GWI], while taking into account frequency of exposure to sarin.

Still, it's intriguing that lower levels of PON1 is there in the PACS fatigue+ group, perhaps suggesting that something is suppressing the body's response to toxins?
(edit - sorry, fixed the last sentence)

 

[SNT Gatchaman]

Listing out the proteins of interest for the search engine:

Reformatting to link these to GeneCards:

Increased:
CFI, LRG1, APOB, IT1H3, CA1, SAA1, IGHV1-2

CFI
LRG1
APOB
IT1H3
CA1
SAA1
IGHV1-2

Decreased:
BTD, LTF, PON1, COMP, IGKVA-1, IGKV1-17, IGHV5-51, GP1BA, IGLV8-61, CFD, MMRN1, PTEN, PROZ, IGHV3-7, RBP4, IGLV1-51, LUM, FN1, PFN1, TAGLN2

BTD
LTF
PON1
COMP
IGKVA-1
IGKV1-17
IGHV5-51
GP1BA
IGLV8-61
CFD
MMRN1
PTEN
PROZ
IGHV3-7
RBP4
IGLV1-51
LUM
FN1
PFN1
TAGLN2

Note fibronectin is down in comparison to Prusty: Increased circulating fibronectin, depletion of natural IgM and heightened EBV, HSV-1 reactivation in ME/CFS and long COVID (2023, Preprint: MedRxiv)

 

[EndME]

Note fibronectin is down in comparison to Prusty: Increased circulating fibronectin, depletion of natural IgM and heightened EBV, HSV-1 reactivation in ME/CFS and long COVID (2023, Preprint: MedRxiv)

Prusty at least partially used a Charite cohort as well (apart from Scheibenbogen being part of both studies), so there might be overlaps in some of the different cohorts. At the time there were also discussions about differences in plasma and serum levels of FN1, as well as FN1 seemingly being less of a measure of anything chronic in that study, rather than a reflection of having been infected with something previously (so chosing a post-mild Covid PACS fatigue+ cohort would seem less relevant with respect to that).

 

[Hutan]

Thanks SNT, very helpful links.

I know googling proteins can be like reading your horoscope in that there's usually something there that feels personally relevant. It's also fun.

Here's a few with high fold change and/or high p values but otherwise randomly chosen.

increased CA1 - from the Wikipedia entry:

CA1 activation is associated with worsened pathological remodeling in human ischemic diabeticcardiomyopathy.[12] In diabetic mellitus type 2 patients with postinfarct heart failure who were undergoing surgical coronary revascularization, myocardial levels of CA1 were sixfold higher than nondiabetic patients. Elevated CA1 expression was mainly localized in the cardiac interstitium and endothelial cells. Furthermore, high glucose-induced elevation of CA1 hampers endothelial cell permeability and determines endothelial cell apoptosis in vitro.[12]

CA1 also mediates hemorrhagic retinal and cerebral vascular permeability through prekallikrein activation and serine protease factor XIIa generation. These phenomena induce proliferative diabetic retinopathy and diabetic macular edema disease progression, which represent leading causes of vision loss.[15]

As CA1 is an important therapeutic target, development of its inhibitors will contribute to disease treatment. Compared to other CA family members, CA1 has relatively low affinity to common CA inhibitors.[16] Nonetheless, it has medium affinity for CA inhibitor sulfonamides.

decreased BTD - biotinidase (note, I don't think the results are suggesting a profound deficiency in the PACS cohort)
Biotinidase Deficiency

Hair loss, skin rash, conjunctivitis, candidiasis, seizures
Ataxia [Ataxia means without coordination. People with ataxia lose muscle control in their arms and legs. This may lead to a lack of balance, coordination, and trouble walking. Ataxia may affect the fingers, hands, arms, legs, body, speech, and even eye movements.]
Hypertonia [decreased muscle tone]

Features of untreated partial biotinidase deficiency may be the same as listed above but are mild and occur only when the person is stressed, such as with a prolonged infection.

Untreated adolescents and adults usually exhibit myelopathy [degeneration of the spinal cord] and optic neuropathy and are often initially diagnosed with multiple sclerosis.

Diagnosis: Detection of deficient biotinidase enzyme activity in serum/plasma

It's worth noting even as just another possible differential diagnosis for FMD and functional seizures - how many people with a functional diagnosis have been tested for BTD?


IG - immunoglobulin proteins
increased - IGHV1-2
decreased - IGKVA-1, IGKV1-17, IGHV5-51, IGLV8-61, IGHV3-7

Immunoglobulin G is the most abundant and the most important Ig in human body, accounting for 70% ~ 80% of the total immunoglobulin, which is a reimmune response antibody.
(1) IgG increase: common in various chronic infections, chronic liver disease, collagen vascular disease, lymphoma and autoimmune diseases such as systemic lupus erythematosus (SLE), rheumatoid arthritis; Simple IgG increase is mainly seen in immunoproliferative diseases, such as IgG-secretory multiple myeloma (MM).
(2) IgG reduction: seen in patients with congenital and acquired humoral immunodeficiency, combined immunodeficiency, heavy chain disease, light chain disease, nephrotic syndrome, viral infection and immunosuppressive agents. It can also be seen in metabolic diseases such as hyperthyroidism and muscular dystrophy.

The IGL stands for the lambda light chains
The IGK stands for the kappa light chains
The IGH stands for the heavy chains

That pattern of some IGH's being decreased and one increased is intriguing, is it just noise? The increased IGHV1-2 might well be, it's only just significant in the volcano plot F (shown above).

 

[Murph]

I have no idea who is right but I like that Prof Scheibenbogen et al are not playing collegiate and politely ignoring a study they disagree with. I would like to see more public fights like this.

 

[Hutan]

I agree that it's nice to see researchers critically looking at studies and having a go to demonstrate the problems. I hope what they have done isn't seen as impolite or fighting though. It's science at work, it's exactly what we need much more of.

Us complaining about cohort matching here on the forum doesn't achieve much. We get better at dismissing findings, but still the misleading papers come. Whereas papers like this, examining the consequences of ignoring cohort differences, will be remembered by young researchers for the rest of their careers, and will help improve their work.

 

[Jonathan Edwards]

That pattern of some IGH's being decreased and one increased is intriguing, is it just noise? The increased IGHV1-2 might well be, it's only just significant in the volcano plot F (shown above).

Ig protein levels vary a lot within normals and Vh gene usage has never turned out to tell us very much except perhaps for Vh4-34 which is anomalous. I suspect noise.