Proteomics of fibrin amyloid microclots in long COVID/post-acute sequelae of COVID-19 (PASC) shows many entrapped pro-inflammatory molecules that may also contribute to a failed fibrinolytic system

Ethical clearance for the study was obtained from the Health Research Ethics Committee (HREC) of Stellenbosch University (South Africa) (references: B21/03/001_COVID-19, project ID: 21,911 (long COVID registry) and N19/03/043, project ID 9521 with yearly re-approval). Participants were either recruited via the long COVID registry or identified from our clinical collaborators’ practice. The experimental objectives, risks, and details were explained to volunteers and informed consent were obtained prior to blood collection. Strict compliance to ethical guidelines and principles Declaration of Helsinki, South African Guidelines for Good Clinical Practice, and Medical Research Council Ethical Guidelines for Research were kept for the duration of the study and for all research protocols.

Blood was collected from 29 volunteers (9 males; 20 females; median age 52 [range 41–57]) to serve as controls. Control patients did not smoke or suffer from coagulopathies, were not pregnant and only 1 volunteer suffered from cardiovascular disease. Those suffering from hypertension or hyperlipidaemia, were well controlled on treatment and none of our controls took anticoagulation medication. Initially, we included 99 long COVID patients in the study (30 males, 69 females.) Liquid chromatography and mass spectrometry were performed on all of the samples and data collected. 66 of the long COVID samples were found to be “high responders” and 24 “low responders” (please refer to the mass spectrometry data analysis section for an in-depth explanation of the classification.) Mass spectrometry data analysis was only carried out on the high responder long COVID samples. The median age of high responders was 51 [40‐60] (21 males; 45 females) and the median age of low responders 45 [31‐58] (5 males; 19 females.) Healthy participants, patients or their attending physicians completed a questionnaire, either as part of the long COVID registry or when they visited their attending clinician. The questionnaire contained information regarding their age, gender, pre-existing comorbid conditions, chronic medication, information regarding the onset, diagnosis, and severity of their acute COVID-19 illness, self-reported long COVID symptoms, as well as their COVID-19 vaccination history (where available).

Stored PPP was used to determine protein concentration using nanodrop technology where a 20 × dilution was made with 10 mM ammonium bicarbonate and protein concentration determined using a nanodrop spectrophotometer (Thermo Fischer) and measuring the absorbance at 280 nm. The samples were then standardised to the lowest protein concentration and 50 µg total protein removed for digestion (25 µL). For the first trypsin digestion step, 50 µg of protein per sample was placed in a tube and 25 µL ammonium bicarbonate and 1 µg of trypsin was added to a final volume of 50 µL. The samples were incubated over night at 37 °C.

For the second trypsin digestion step, the samples were centrifuged the next day at 16,000g for 10 min (volume in each sample were still 50 µL). Next, 42 µL of supernatant were removed carefully to not do not disturb the microclots at bottom of tube. To the remaining 8 µL, 10 µL of 250 mM Triethyamonium bicarbonate (TEAB) (ThermoFisher) containing 5 mM tris carboxyethyl phosphine (TCEP) (Sigma), was added. The samples were vortexed for 5 s followed by incubation of 10 min at 90 °C. Samples where then incubated for a further 50 min at 60 °C, followed by a cooling-down step. Cysteine residues were blocked using 10 mM (final concentration) methyl methanethiosulfonate (MMTS) (we added 1 µl into each sample). Samples were left at room temp for 30 min. Thereafter, 20 µL of 50% (Sigma) methanol was added and samples were vortexed for 5 s (to denature last intact proteins that might remain). Now 10 µL of trypsin solution in 100 mM TEAB (containing 1 µg of trypsin) (Thermo) was added to each sample followed by incubation for 18 h at 37 °C. Samples were then acidified with 5 µL trifluoroacetic acid (TFA) (Sigma) of which the final concentration of 0.5% was used to precipitate trypsin. Samples where centrifuged at 16,000g to pellet the trypsin precipitate and the supernatant removed and placed into HPLC inserts. Samples were allowed to evaporate to dryness under vacuum (roto-evaporator) followed by resuspending in 20µL loading buffer consisting of 2% acetonitrile (Burdick and Jackson) in deionized water containing 0.1% Formic acid (Sigma) spiked with Biognosys 11 indexed retention times standards (Biognosys.) The dissolved samples were loaded directly in the autosampler set to 4C.

Liquid chromatography was performed on a Thermo Scientific Ultimate 3000 RSLC [23, 24] equipped with a 20 mm × 100 µm C₁₈ trap column (Thermo Scientific) and a CSH 25cmx75µm 1.7 µm particle size C₁₈ column (Waters) analytical column. The solvent system employed was loading: 2% acetonitrile:water; 0.1% FA; Solvent A: water; 0.1% FA and Solvent B: 100% acetonitrile, 0.15% FA. Methods were previously described in [1]. Separation was effected using a linear gradient from 2 B to 30% B over 65 min followed by 30 B to 45% B from 65 to 80 min. The sample was loaded onto a trap column at 2 µL/min before switching to the analytical column at three minutes with a flow rate of 300 nL/min. Chromatography was performed at 45 °C and the outflow delivered to the mass spectrometer through a stainless-steel nano-bore emitter.

Mass spectrometry was performed using a Thermo Scientific Fusion mass spectrometer equipped with a Nanospray Flex ionization source. Data was collected in positive ion mode with spray voltage set to 1.8 kV and ion transfer capillary set to 275 °C. The mass spectrometer auto-calibration was activated using polysiloxane ions at m/z = 445.12003. MS¹ scans were performed using the orbitrap detector set at 120,000 resolution over the scan range 375–1500 with AGC target at 2 E⁴ and maximum injection time of 50 ms. Data were acquired in profile mode. MS² acquisitions were performed using monoisotopic precursor selection for ion with charges + 2 to + 7  with error tolerance set to ± 10 ppm. Precursor ions were excluded from fragmentation once for a period of 60 s. Precursor ions were selected for fragmentation in HCD mode using the quadrupole mass analyser with HCD energy set to 30%. Quadrupole isolation was set at m/z = 1.2. Fragment ions were detected in the Orbitrap mass analyser set to 30,000 resolution. The AGC target was set to 5E4 and the maximum injection time to 100 ms. The data was acquired in centroid mode. Samples were randomised by using a randomization schedule. Technical replicates are not feasible in an experiment with a large sample size such as this.

The raw files generated by the mass spectrometer were imported into SearchGUI [25] and processed using the MsGF + algorithm. Database interrogation was performed against the 2019-nCOVpFASTA database [26]. Semi-tryptic cleavage with 2 missed cleavages was allowed for. Precursor mass tolerance was set to 10 ppm and fragment mass tolerance set to 0.02 Da. Deamidation (NQ), oxidation (M) and Phosho-ST were allowed as dynamic modifications and methylthio of C as a fixed modification. Data was imported into Skyline and the extracted ion chromatograms of positively identified peptides summated as indication of relative protein abundance. Spectral data were evaluated in Skyline and the proteomics sample was divided into high and low responders. Samples where 90% of all the protein intensity level read outs (or the area under the peak) were above the 3rd quartile when compared to the protein levels observed in our other samples, were classified as “high responders". If 90% of all the protein intensity level read outs were below the 1st quartile when compared to the proteins in other samples, they were classified as “low responders.” For example, if there were 100 proteins identified but 90 of those proteins had concentrations/responses that were below the 1st quartile compared to the other samples in the sample set, the patient was classified as a low responder. Low responders were excluded from further analysis. We ensured that it was not attributable to instrument performance and batch effects were also excluded. The search results were imported into Scaffold Q + for further validation (www.​proteomesoftware​.​com) and statistical testing. The (FDR) false discovery rate was set to 1%. Normalization was selected in Scaffold Q + with a weighted spectra quantitation method and missing values were processed by the algorithm contained within the software. A Fischer’s exact test (F-test) was performed on the dataset and Benjaminin-Hochberg multiple testing correction applied. This resulted in a p-value of < 0.012. For immunoglobulin specific searches the FASTA files were obtained from uniprot (www.​uniprot.​org) by searching the database for human immunoglobulins and downloading the returned protein sequences. Protein parsimony for immunoglobulin searches was handled by peptide and Protein Prohpet algorithms within Scaffold. The immunoglobulins listed later in the paper (as shown in Table 3 and Table 4,) were cross referenced on THE INTERNATIONAL IMMUNOGENETICS INFORMATION SYSTEM®’s IMGT/LIGM-DB database (nucleotide sequences of IG and TR from 360 species.) The citable accession numbers of the immunoglobulins were used to determine the EMBLⁱ reference number to search for the particular immunoglobulin on the database: https://​www.​imgt.​org/​ligmdb/​ Our proteomics data will be deposited in an international repository such as PRIDE (EBI) for public access.

Fluorescence microscopy of microclot formation in PPP was performed on some of our samples as described in our previous papers [17]. PPP were exposed to the fluorescent amyloid dye, Thioflavin T (ThT) (final concentration: 0.005 mM) (Sigma-Aldrich, St. Louis, MO, USA) for 30 min (protected from light) at room temperature [18, 27‐30]. After incubation, 3 µL stained PPP was placed on a glass slide and covered with a coverslip. The excitation wavelength band for ThT was set at 450 to 488 nm and the emission at 499 to 529 nm and processed samples were viewed using a Zeiss Axio Observer 7 fluorescent microscope with a Plan-Apochromat 63x/1.4 Oil DIC M27 objective (Carl Zeiss Microscopy, Munich, Germany).

Haematocrit samples of a few of our patients in the cohort were exposed to the two fluorescent markers, CD62P (PE-conjugated) (platelet surface P-selectin) (IM1759U, Beckman Coulter, Brea, CA, USA) and PAC-1 (FITC-conjugated) (340507, BD Biosciences, San Jose, CA, USA) (17). CD62P is a marker for P-selectin that is either found on the membrane of platelets or inside them. PAC-1 identifies platelets through marking the glycoprotein IIb/IIIa (gpIIb/IIIa) on the platelet membrane. To study platelet pathology, 4 µL CD62P and 4 µL PAC-1 was added to 20 µL haematocrit, followed by incubation for 30 min and protected from light at room temperature. The excitation wavelength band for PAC-1 was set at 450 to 488 nm and the emission at 499 to 529 nm and for the CD62P marker it was 540 to 570 nm and the emission 577 to 607 nm [17]. Samples were viewed using a Zeiss Axio Observer 7 fluorescent microscope with a Plan-Apochromat 63x/1.4 Oil DIC M27 objective (Carl Zeiss Microscopy, Munich, Germany).

Statistical analysis was done using Graphpad Prism 8 (version 8.4.3). All data were subjected to Shapiro-Wilks normality tests. An unpaired T-test was performed on parametric data with the data expressed as mean ± standard deviation, whereas the Mann–Whitney U test was used on unpaired non-parametric data and the data expressed as median [Q1–Q3] (all two-tailed).

We found increased levels of molecules that may modulate coagulation (see Fig. 4 for the areas in the clotting cascade where these molecules will have physiological activity). The most significant ones were VWF and PF4. VWF is a well-known molecule in vascular (endothelial) injury, platelet-platelet adhesion and is generally involved in hypercoagulation, if found to be increased [11, 50] which could contribute to coagulopathies seen in patients with long COVID. PF4 is a small cytokine belonging to the CXC chemokine family that is also known as chemokine (C-X-C motif) ligand 4 (CXCL4). Platelets store PF4, and it is released from the platelet during activation [51]. Heparin can bind to PF4 and promotes PF4 aggregation, resulting in the formation of PF4/heparin complexes, that have antigenic properties [51]. Recently, it was reported that COVID-19 patients might present with reactivity in PF4/heparin antigen tests without the presence of platelet-activating antibodies [52]. PF4 and VWF may also form complexes; recognition of PF4-VWF complexes by heparin-induced thrombocytopenia antibodies contributes to thrombus propagation [53]. Recently, in a smaller cohort of samples, we found that α-2AP was increased in long COVID samples [1]. In the current analysis, it was marginally increased. Of significant importance, is that platelets are significantly hyperactivated in long COVID patients. This is noted in both microscopy (as shown in Fig. 2) and the proteomics analysis. Also, it is well-known that individuals with long COVID struggle with anxiety [54, 55]. This was also seen in our long COVID cohort, where 30% of the participants reported anxiety as a symptom (see Table 1). Platelets are well-known for their storage of serotonin (5-HT) [56] and PF4 and serotonin are stored in α- and δ-granules [57]. Circulating immune complexes may also activate platelets via receptor-receptor binding followed by the release of serotonin from platelet granules [57]. Hyperactivated platelets may shed serotonin (and other molecules stored inside platelets), resulting in depletion of serotonin and other molecules stored inside of platelets. The serotonin transporter (SERT or 5HTT) (https://​www.​ncbi.​nlm.​nih.​gov/​books/​NBK53687/​) is expressed on the plasma membrane of several bodily tissues, for example blood platelets [58, 59], the brain, lungs, bone, gastrointestinal tract as well as the cardiovascular system and transports the neurotransmitter serotonin (5-hydroxytryptamine) across the cell membrane. Prior research suggested that the transport of serotonin (5-HT) via the SERT as well as the storage, metabolism and release mechanisms thereof are similar in platelets and serotonergic neurons. Serotonin depletion may be a significant contributor to the depressed mood or anxiety noted in individuals suffering from long COVID.

Plasma kallikrein (KLKB1) levels were found to be decreased in the current long COVID cohort, and is known to activate factor XII, and it also belongs to a subgroup of trypsin-like serine proteases, enzymes capable of cleaving peptide bonds in proteins [60]. It is also involved in the bradykinin pathway, that is seen as an inflammatory product of the coagulation system [61]. Both neutrophil and mast-cell activation are functionally linked to the bradykinin pathways [61], and mast cell activation is a known feature of PASC, consistent with the known benefits of antihistamine treatment [62] KLKB1 also digests plasminogen to plasmin and participates in surface-dependent activation of blood coagulation, fibrinolysis, and inflammation. Under physiological conditions, plasma kallikrein serves as a cardioprotective enzyme, but when it is increased in circulation, it is involved in cardiovascular disease [63]. A decrease in the level of this molecule may prevent adequate plasminogen activation and may directly impact downstream on clot breakdown and therefore fibrinolysis (see Fig. 5.)

Molecules that modulate cellular function that we found to be increased, included galectin-3-binding protein, Thrombospondin-1, Alpha-1-acid glycoprotein-2 as well as Inter-alpha-trypsin inhibitor heavy chain H1 and 2 (ITIH1 and ITIH2). Galectin-3 expression is known to be induced in viral infections and by a multitude of molecules that either mimic or are characteristic of ongoing inflammation and microbial infection, such as IFN-α, IFN-β, IFN-γ, TNF-α, poly(I:C), dsRNA, and dsDNA [64]. In addition, it is commonly overexpressed in most types of cancers [65] and is an important player in cancer development, progression, and metastasis via its interactions with galactoside-terminated glycans [66]. Furthermore, it interacts with the cell-surface glycoprotein CD146 (MCAM, MUC18) and induces secretion of metastasis-promoting cytokines from vascular endothelial cells [66]. Recently, it was also identified as a novel predictor of venous thromboembolism in systemic lupus erythematosus [67].

Thrombospondin-1 was also found to be increased in our cohort of long COVID participants. It is an adhesive glycoprotein that is released from platelet alpha-granules in response to thrombin stimulation and is also a transient component of the extracellular matrix in developing and repairing tissues [68]. Recently it was suggested that it might regulate haemostasis in vivo through modulation of platelet cAMP signalling at sites of vascular injury [69], and that it could mark a prothrombotic state and an underlying inflammatory process [70]. It is also increased in the tumour microenvironment [64] and might be involved in antiphospholipid syndrome [70], a condition well known to be associated with thrombosis.

Alpha-1-acid glycoprotein-2, which we also found to be elevated, is known to modulate the immune system during the acute-phase reaction. ITIH1 and ITIH2 were found to be elevated in our long COVID population. These molecules belong to the inter-α-trypsin inhibitors. Previously, it was reported that ITIH1 and ITIH2 were more abundant in a COVID-19 survivor group [71]. ITIH1 has previously been found to be elevated in patients with osteoarthritis [72]. ITIH1 is synthesized by chondrocytes and binds to hyaluronic acid and also to the extracellular matrix [73]. ITIH2 has been associated with demyelinating diseases [74].

We found decreased levels of a few molecules, namely Long palate, lung and nasal epithelium carcinoma-associated protein 1 (LPLC1_HUMAN,) Lactotransferrin, Adiponectin and Alpha-1-acid glycoprotein 1 (A1AG1_HUMAN) that might point to a type of immunosuppression. Our research group also suggested that Lactoferrin may be beneficial for preventing adverse effects of acute COVID-19, mainly due to its antiviral properties [75]. Previously, it was found that some critically ill COVID-19 patients could not be discharged from the ICU even though they exhibited negative viral tests [76]. Zhou and co-workers in 2021 suggested that in such patients, adaptive immunosuppression might be present, and ascribed it to a dysregulated host response to severe infection, similar to the immunosuppression that exists in patients with sepsis [76]. In long COVID, there has also been reports of immune suppression [54].

Of importance may be LPLC1, which may play an important role in innate immunity, as its function is related to binding of LPS, as a LPS-binding protein (https://​www.​uniprot.​org/​uniprot/​C3TTP7). If LPS is not bound effectively, it may have significant implications for the response to LPS in circulation. This could possibly indicate an altered immune response to infectious pathogens. Related to LPS binding, is Lactotransferrin also known as lactoferrin (LF,) that was also decreased in our long COVID cohort. LF has important immunological properties and is both antibacterial and antiviral. In particular, there is evidence that it can bind to at least some of the receptors used by coronaviruses and thereby block their entry [75]. LF makes use of Heparan Sulfate Proteoglycans (HSPGs) on the surface of epithelial cells of the stomach to facilitate entry into the bloodstream via endocytosis [77, 78]. Coronaviruses bind to host cells by first attaching to HSPGs utilizing them as preliminary docking sites to assist with viral entry and LF is able to interfere with some of the receptors used by coronaviruses. LF can inhibit the entry of viral particles into host cells by blocking cellular receptors or via direct attachment to viral particles. In COVID-19 infection, LF might be able to competitively bind receptors as well as HSPGs and therefore prevent viral entry and the subsequent cytokine storm [75] but this still requires further investigation. Our finding of decreased levels of lactoferrin possibly indicates suppression of the host immune response in patients with long COVID.

Adiponectin has anti-inflammatory and anti-apoptotic properties that has shown to protect the vasculature, heart, lung, and colon [79], and was also reduced in our long COVID population. Pro-inflammatory and anti-inflammatory cytokines or adipokines namely adiponectin, leptin and resistin may be released from adipose tissue [80]. Adiponectin exists in two isoforms namely HMW (high molecular weight) and LMW (low molecular weight) adiponectin and is an important regulator of cytokine immune responses. These cytokine responses are isoform specific. IL-6 secretion by human monocytes and human monocytic leukemia cell line cells are stimulated by HMW adiponectin and does not inhibit LPS-induced IL-6 secretion. On the other hand, LMW adiponectin promotes IL-10 secretion and inhibits LPS-mediated IL-6 release. For example, in chronic hepatitis C infection these cytokines play an important role in the inflammatory and immune response of affected individuals. This could explain the different outcomes seen in hepatitis C infection. Some of the metabolic and inflammatory effects of adiponectin are regulated through the mitogen-activated protein kinase (MAPK) signalling pathway [80]. The MAPK pathway forms a crucial part in the regulation of IFN-α and INF-γ signalling in viral infections such as hepatitis C. The reduced level of adiponectin in our long COVID cohort also potentially points to an altered immune response in infected patients.

Alpha-1-acid glycoprotein 1 is an acute-phase protein [81] and functions as a transport protein in the blood stream, and it was found to be reduced in our long COVID participants. It is known to regulate inflammation, including binding of pathogens and modulating white blood cells’ activity throughout the entire leukocyte attacking sequence [82]. This molecule is also known to modulate the immune system during the acute-phase reaction and decreased levels are related to the mortality of septic patients [83]. The decreased levels seen in the current cohort, might possibly indicate a dysregulated or abnormal immune or inflammatory response. It may therefore also be a predictor of morbidity and mortality in patients with COVID-19. If we look at the physiological processes that are regulated by the aforementioned molecules, the increased or reduced levels of these molecules will have an effect on cellular function and the potentially dysregulated immune response in these patients. They also have immune-related or cellular remodelling properties and may impact on vascular function.

We also found interesting apolipoproteins, known to play a role in lipid metabolism and regulation. Apolipoprotein C-II (a component of very low density lipoproteins and chylomicrons) was found to be increased while Apolipoprotein A-II (the second most abundant protein of the high density lipoprotein particles) was reduced in our cohort. Previous studies support the crucial role that lipids and lipid metabolism play in SARS CoV-2 infection [84‐86]. Shen et al. performed proteomic and metabolomic profiling on the serum of 46 COVID-19 patients compared to 53 controls [84]. They found the downregulation of over 100 lipids, including sphingolipids, glycerophospholipids and fatty acids which correlated with disease severity in patients with COVID-19. Lipids, including sphingolipids and glycerophospholipids form an essential part of biomembranes. Specialized membrane domains exist in the plasma membrane and sphingolipids are an example of this. Sphingolipids are also involved in the regulation of signal transduction, immune activation pathways and inflammatory responses [87]. The downregulation of lipids in the sera of patients suffering from COVID-19 points to liver damage and changes in bilirubin and bile acids supports this. Song et al. studied the lipidome and metabolome of 50 patients with different severities (mild, moderate, and severe) of SARS CoV2 infection compared to 26 healthy controls by utilizing targeted and untargeted mass spectrometry of plasma samples that were collected [85]. They noted reductions of major classes of plasma glycerophospholipids, phosphatidic acids (PAs), phosphatidylinositols (PIs), and phosphatidylcholines (PCs) and increased concentrations of lysophospholipids namely lysophosphatidic acids (LPAs,) lysophosphatidylinositols (LPIs,) and lysophosphatidylcholines (LPCs.) Another finding was that the plasma lipidome of patients with COVID19 had similarities to monosialodihexosyl ganglioside (GM3)-enriched exosomes where levels of sphingomyelins (SMs) and GM3s are elevated, and diacylglycerols (DAGs) decreased. Increasing levels of GM3s were seen with progressive disease severity indicating that GM3-rich exosomes also play a role in the pathophysiology of COVID19.

From the literature it is evident that fluctuations in DGs (diglycerides,) FFAs (free fatty acids), and TGs (triglycerides) occur in pathological conditions. For example, increased lipolysis of adipose tissue is seen in Ebola virus disease where TGs are converted to FFAs and DGs, with increased recycling of fatty acids back into TGs [88, 89]. Bruzzone et al. found elevated levels of serum TG, TG-VLDL (VLDL, very low-density lipoproteins), TG-IDL (IDL, intermediate density lipoproteins), TG-LDL (LDL, low-density lipoproteins) and TG-HDL (HDL, high density lipoproteins) in patients with COVID-19 in contrast to a decreased total cholesterol (TC) concentration. They propose that serum accumulation of TG and TG-VLDL may indicate decreased oxidation of acetyl-CoA in the mitochondria of the liver. Ketone bodies are produced by the liver from fatty acid oxidation-derived acetyl-CoA [90]. This pathogenic redistribution of the lipoprotein particle size and composition increases the cardiovascular risk for patients with COVID-19 infection.

Literature supports the role that apolipoproteins and steroid hormones have to play in SARS CoV-2 infection. Shen and co-workers in 2020 [84] found decreased levels of apolipoproteins APOA1, APOA2, APOH, APOL1, APOD, and APOM in the serum of patients following SARS CoV2 infection. It has also previously been reported that APOA1 levels decreased when patients progressed from mild to severe disease [91]. These apolipoproteins are involved in the functioning of macrophages [92]. Arachidonic acid (AA) is an essential fatty acid. Prostaglandins, leukotrienes and thromboxanes are pro-inflammatory metabolites of AA and eicosapentaenoic acid (EPA). Lipoxins, resolvins, protectins and maresins are derivates of AA, EPA and docosahexaenoic acid (DHA) and are able to suppress the inflammatory response, enhance phagocytosis of macrophages as well as other immune cells and assist with microbial clearance [92, 93]. Therefore, lipids such as glycerophospholipids and fatty acids play a pivotal role in inflammation and the host immune response against viral infections.

In summary, lipid metabolism plays an integral role in COVID-19 disease, resulting in an imbalance between immune-regulating lipids, pro-inflammatory lipids and lipid mediators during the host immune response. It is evident that our findings of increased levels of Apolipoprotein C-II and reduced levels of Apolipoprotein A-II also supports this dysregulation of lipid metabolism in patients suffering from COVID-19 (see Fig. 5).

Table 3 shows selected immunoglobulins involved in the immune response and that might act as antibodies or autoantibodies. As can be noted in Table 3, there are numerous molecules with no data available on their function. However, we do list these molecules here, as it may be of value for immunologists to note the presence of these potentially important antibodies. Of specific interest was the presence of Immunoglobulin kappa chain V-III region POM (plasma membrane), that is known as rheumatoid factor (RF) and that may act as an autoantibody [33]. Another RF factor that was found to be elevated was Immunoglobulin kappa chain V-III region CLL.

Of significance also, was that there were numerous antibodies only present in 63 of our long COVID high responder samples and we present those in Table 4 that were present in 13% or more (arbitrary chosen cut-off) of the cohort. IG c1641_light_IGKV1-6_IGKJ4 (Fragment), IGL c3450_light_IGKV1-39_IGKJ1 (Fragment) and IGH c2558_heavy_IGHV3-20_IGHD2-2_IGHJ2 (Fragment) were found to be present in 27%, 24% and 22% of the 63 samples respectively. These antibodies were cross referenced on THE INTERNATIONAL IMMUNOGENETICS INFORMATION SYSTEM®’s IMGT/LIGM-DB database, but no data are available on these antibodies. These antibodies may have significance in the immune response following acute COVID-19 illness or driving auto-immunity in some long COVID participants.

Another important consideration is that patients with long-term persistent microclots, may also progress to develop auto-immune diseases. We suggest that this phenomenon might directly be due to autoantibodies that develop against antibodies or other molecules that are entrapped inside the fibrinaloid microclots. It appears that the long-term sequelae of many viral infections [94, 95], including long COVID [96, 97] may be attributed to autoimmunity. These microclots contain many different proteins that may present new epitopes if a conformational change has taken place. These epitopes may be recognized as ‘non-self’ eliciting an immune response with the possible development of autoantibodies. Additionally, oxidative stress induces the nitration of proteins leading to the production of autoantibodies. Autoantibodies formed against actin lead to muscle weakness [98]. In ME/CFS (Myalgic Encephalomyelitis/Chronic Fatigue Syndrome), autoimmunity is well described [94]; and might also play an important role in long COVID because some autoantibodies share elements of epitope with the SARS-CoV-2 virus [99]. It has also been suggested that some of the viral sequences are amyloidogenic [100, 101]. Viral persistence may play a key role in patients suffering from long COVID [102], and therefore driving an ongoing immune response or the development of autoimmunity (see Fig. 5).

Pascolini et al. studied 33 consecutive patients with COVID-19 of whom 94% had interstitial pneumonia and compared them to 25 age- and sex-matched controls with fever and/or pneumonia of an etiology other than COVID-19 to evaluate for autoimmune serological markers [103]. 45% of patients with COVID-19 tested positive for at least one autoantibody and they found that 33% tested positive for antinuclear antibodies (ANAs), 24% tested positive for anti-cardiolipin antibodies (immunoglobulin (Ig)G and/or IgM) and 9% tested positive for anti-β2-glycoprotein antibodies (IgG and/or IgM.) Antineutrophil cytoplasmic antibodies (ANCA) were not present in any of their patients. A higher mortality rate was seen amongst patients with autoantibodies compared to those without.

Chang and coworkers searched for autoantibodies in hospitalized patients with COVID-19 and found that approximately 50% of hospitalized patients developed autoantibodies against one or more antigens in the array they tested for with 25% of patients being ANA positive [104]. They also discovered antibodies that are associated with relatively rare connective tissue diseases and have been predicted to be pathogenic. A large number of anti-cytokine antibodies (ACA) were identified as well as antibodies recognizing non-structural SARS-CoV-2 proteins which correlate with autoantibodies. A few of the antibodies they discovered include Anti-C1q, which is a SLE autoantigen, anti-β2GP1 (anti-beta 2 glycoprotein 1) a thrombogenic antibody, anti-BPI (anti-bactericidal permeability inducing protein) and anti-ACE2 antibodies.

Conversely, a study performed by Klein et al. suggest that autoantibodies have a limited role to play in the pathogenesis of long COVID [105]. Their findings point to the involvement of persistent viral antigen, reactivation of latent herpesviruses, and chronic inflammation in long COVID pathophysiology. There have been previous reports of persistent viral antigen in intestinal biopsies of patients recovering from COVID-19 infection and antigen persistence may therefore contribute to the host  immune response in patients with long COVID [106, 107].