As mentioned in the introduction, several PRRs sense DNA (see Figure 1 ) and, with exception of TLR9, they do so in a sequence-independent manner. This means that in principle, these PRRs do not distinguish self- from nonself-DNA. This lack of a specificity for any sequence motifs explains why cGAS is now considered one of the elements that connect DNA damage to several autoimmune diseases and cancer, but also the counterintuitive observation of cGAS-controlled immunity against RNA viruses, including HIV-1, human T cell-leukaemia virus type I and Dengue virus (172–174). Active cGAS produces cyclic GMP-AMP to activate STING, which dimerises and activates TNF receptor associated factor (TRAF) associated NF-κB activator (TANK) binding kinase 1 (TBK1) and downstream, interferon regulatory factor 3 (IRF3), which moves into the nucleus and triggers the transcription of type I and type III interferon (IFN). In the particular case of SARS-CoV-2, the cell-autonomous activation of cGAS/STING signalling also contributes to NF-κB-dependent cytokine production, at least in human epithelial cells (15).

Besides cytosolic cGAS, endosomal TLR9 has been reported to be activated by SARS-CoV-2 (18, 175). TLRs are membrane-bound PRRs that sense a diverse array of extracellular or endosomal DAMPs and PAMPs and form part of one of the most studied examples of a DAMP-mediated immune priming: the activation of inflammasomes in two steps (59, 176, 177). Upon binding by one of these ligands, most TLRs signal via myeloid differentiation primary response (MyD)88, Interferon Receptor-Associated Kinases (IRAKs), TNF receptor-associated factor (TRAF)6, TBK1, and the IKK complex (69, 178). Downstream, activated NF-κB translocates to the nucleus to facilitate the expression of pro-IL-1β, IL-6 and pro-IL-18 and of the different elements of the inflammasome: AIM2, NLRP3, apoptosis-associated speck containing a caspase recruitment domain (CARD) (ASC)-like protein, pro-caspase, and pro-Gasdermin D ( Figure 1A ) (40, 58, 179). Thereby, TLR ligands serve as signal 1 that primes the cell for fast responses to future, more challenging threats, which are indicated by a second, intracellular signal 2. Both AIM2 and NLRP3 are dsDNA sensors that bind to cytoplasmatic DNA, including phagocytosed DNA that is released from lysosomes. AIM2 seems to exclusively sense DNA, but NLRP3 senses multiple endogenous and exogenous molecules that indicate threats, including reactive oxygen species (ROS), and it is also activated when extracellular ATP (eATP) binds to a purinergic receptor, e.g. P2X7 (38, 59, 180). In response to this ‘activation’ signal, AIM2 or NLRP3 associate with ASC and pro-caspase 1 to form the active inflammasome, a multiprotein complex that liberates active caspase-1 to facilitate the maturation of IL-1β and IL-18. Since these ILs don´t possess a secretion signal, active gasdermin D is also produced to form a plasma membrane pore, which enables the release of these cytokines together with cellular content (13, 181) ( Figure 1B ).

In the context of COVID-19, the dsDNA sensors participate in inflammatory processes that usually culminate in cell death, via mechanisms that can be roughly grouped into three major pathways: Cell-autonomous cell death, bystander cell death and dying syncytia. Cell-autonomous cell death leading to aberrant inflammation in COVID-19 ( Figure 2A ) has been described, e.g., by Andrea Ablasser’s team in Lausanne (9) and Ralf Bartenschlager’s team in Heidelberg (15) as an outcome of mtDNA release from damaged mitochondria that results in an ‘aberrant’ cGAS/STING-pathway causing a specific activation of NF-κB and a block of IRF3 nuclear translocation in SARS-CoV-2 infected endothelial cells and lung epithelial cells, and by a Brazilian team with first author Tiago Costa (18) as the outcome of a mitochondrial dysfunction that leads to a TLR9-dependent NF-κB activation in human umbilical vein endothelial cells ( Figure 2A ). Furthermore, a team at NIH in Bethesda and Georgetown University in Washington (29) described autonomous cell death to be caused NLRP3-inflammasomes activation by high levels of oxidative stress that are associated with mitochondrial dysfunction in human monocytes (29), and also Judy Lieberman’s and Richard Flavell’s teams at Harvard Medical School and Yale (22, 28) inflammasome-dependent pyroptosis of SARS-CoV-2 infected macrophages and monocytes ( Figure 2B ). Doubts remain, and several differences among these papers will have to be resolved by future work. For example, Junqueira et al. observed ca.10 % infected cells among the blood monocytes in COVID-19 patients (22), while Lage et al. “were unable to detect productive infection of primary human HC monocytes by SARS-CoV-2 in vitro” (29). Di Domizio et al. (9) define endothelial cell death as STING-dependent but do not offer a concrete mechanism that connects aberrant IFN I-signalling to the death of infected cells and – curiously enough – I could not find a single study that considered both cGAS/STING signalling and inflammasomes together. Doubts also remain concerning the signals I and II that are generated by SARS-CoV-2 replication to activate inflammasomes in the same cell and whether activation of TLR9 was caused by mtDNA or a direct effect of the virus. Both mitochondrial ROS (29) and DNA from other dying cells (Lieberman, personal communication) were suggested to activate inflammasomes as signal II, which in the latter case would classify the cell death as bystander cell death.

Evidently, infected cells eventually die. Before or after doing so, they can trigger inflammation and eventually, cell death, in neighbouring immune ‘bystander’ cells. The team headed by Andrea Ablasser (9) observed that activated, IFN I-producing macrophages frequently surrounded those vessels that exhibited strongest signals of endotheliopathy and concluded that “signals derived from dying (endothelial) cells promote type I IFN production by macrophages”. Similarly, a particularly nice and detailed study performed in Jenny Ting’s Lab in Chapel Hill and published by Katherine Barnett as first author (13) started from immunohistochemical analyses of COVID-19 autopsy lungs, which showed active inflammasomes and cell death in alveolar macrophages directly adjacent to infected alveolar epithelial cells. Subsequently, they used a co-culture system of human airway epithelial cells and peripheral blood mononuclear cells to confirm that inflammasome activation and cell death was limited to co-cultured cells but absent in isolated cells. Both studies suggested self-DNA (both mitochondrial and genomic) from the dying infected cells as the signal that activated inflammation in the immune cells, although Di’Domizio et al. favoured engulfed endothelial cells ( Figure 2C ) while Barnett et al. suggest lytic cell death as the mechanism allowing bystander cells to access DNA from infected cells (9, 13).

Third, SARS-CoV-2-infected cells can fuse with non-infected ACE2-expressing cells to form syncytia: multinucleated cell complexes that facilitate viral spread without exposure to host antibodies (23, 182, 183) ( Figure 3 ). Various groups observed syncytia as a common cytological feature of post-mortem lung samples obtained from individuals who died of COVID-19 (23, 182, 184–186). A team from Beijing and Harvard with first author Zhou Zhuo (17) focused on the switch from a suppression of type I IFN signalling during the initial phase of infection to the (over-) induced cytokine signalling at later stages and discovered that this shift is associated with syncytia formation, which in turn is accompanied by the release of chromatin – including nDNA to the cytosol (13, 17). Thereby, SARS-CoV 2 enhances the visibility of its’ hosts self-DNA to the immune system (187).

Pharmacological attempts to support a causal role of DNA in COVID-19 used DNA-scavenging nanomaterials or recombinant DNase-I (Dornase α). Bruce Sullenger’s lab at Duke University discovered that the plasma and endotracheal aspirate of COVID-19 ICU patients activated various TLRs, including TLR9. Aiming at identifying DNA-containing DAMPs as these agonists, they treated serum and ETA with DNA-scavenging MnO nanoparticles and observed strong reductions of the content of DNA, RNA and HMGB-1 and consequently, the activating effect on TLRs (14). Unfortunately, the hypothesised ‘DNA-containing DAMPs’ were not identified and serum, rather than individual DAMPs, was used in all TLR-induction tests. Although DNA/HMGB1 or DNA/histone complexes were likely contributing to the fraction of DNA-containing DAMPs, this study clearly shows that much more research will be required until certain inflammatory effects can be clearly related to specific DAMPs.

Several groups treated COVID-19 patients with recombinant human DNase I, administered as DNase I-coated nanospheres or as nebulized Dornase α, which is approved for the treatment of cystic fibrosis (117, 119, 122). I already mentioned the team from the University of Missouri (116). Subsequently, Hacer Kuzu Okur and colleagues at Acibadem Altunizade Hospital in Istambul (117) reported on 7^th of September 2020 that treatment of patients with Dornase α lead to significant clinical improvement in the radiological analysis, oxygen saturation and respiratory rate. These changes were associated with significantly decreased viral loads when comparing nasopharyngeal and oropharyngeal samples taken on the day before the treatment and after 7 days. In the same study, Dornase α decreased viral load and the negative effects of SARS-CoV-2 infection on cell proliferation in a realtime Vero cell culture system as well as in vitro NETosis by thawed adult human mononuclear cells (117). Three weeks later, Andrew G. Weber and coauthors from New York (118) reported on significant reductions in the production of proinflammatory cytokines by PBMCs and in the fraction of inspired oxygen in five mechanically ventilated COVID-19-patients who were treated with nebulized Dornase α and who were successfully extubated, discharged from hospital and remained alive (118). A further month later (on 20^th of October 2020), a multicentre team from S. Korea with first author Hee Ho Park (119) reported that treating plasma of COVID-19 patients with DNase-I significantly reduced the eDNA levels and NET formation. Moreover, DNase-I-coated nanospheres decreased eDNA concentrations in the blood, neutrophil activity, lung damage and mortality in a septic mouse model (119).

A team from Sweden with first author Jane Fisher (120) treated five severely ill COVID-19 patients with Dornase α and all of them became independent of mechanical ventilation and could be dismissed from the intensive care unit within 4 to 15 days of treatment. Immunofluorescence microscopy of sputum produced by these patients confirmed the infiltration by neutrophils and high abundances of NET-forming neutrophils in COVID-19 sputum, and ex-vivo treatment of the sputum from one patient demonstrated the rapid (within 10 min) degradation of NETs. In the same study, proteomic analyses of sputum allowed to identify both subtypes of immunoglobulin A and mucins, blood plasma proteins such as albumin, leukocyte proteins, and inflammatory/antiviral response proteins such as interferon-induced proteins as the most abundant proteins in COVID-19 and to reveal that the recovery due to Dornase α treatment was associated with a reduction in complement proteins, haemoglobin, lipopolysaccharide protein, and C-reactive protein, that is, proteins indicative of elevated innate immunity, cell damage and ongoing infection (120). In a study conducted by Andrew Toma and co-authors from Palm Beach Gardens Medical Centre (121), of 39 patients included in the study, 24 had reduced respiratory support requirements and out of 8 patients initially requiring mechanical ventilation, 7 were successfully extubated by the end of the study period.

A Greek team with first author Efstratios Gavriilidis (122) combined inhaled DNase with drugs aimed at controlling cytokine-mediated hyperinflammation: an inhibitor of the IL-6-receptor and a Janus Kinase (JAK)1/2 inhibitor. They report that the combined treatment was associated with significantly lower in-hospital mortality and intubation rate, shorter duration of hospitalization, and prolonged overall survival. As a notion of caution, in a clinical trial performed by the team at University of Missouri with 10 COVID-patients treated with Dornase α and 20 case controls and published 2021 with first author Zachary Holliday (188), the differences between treatment and control groups in static lung compliance and NET abundance in BALF were not maintained after terminating the treatment. The same study could detect on effect on NETs in the blood (188). In summary, a positive effect of Dornase α has so far only been shown for NETs on the surface, but not in the blood vessels of human COVID-19 patients.

Additional, although indirect support for the detrimental effects of DNA-driven inflammation is provided by the fascinating explanation as to why coronavirus infection does not cause disease symptoms in bats (189–193). The energy-demanding flight of bats unavoidably damages their mtDNA (191, 192). As an adaptation, bats have lost the inflammasome-forming PRRs AIM2 and IFI 16 and other IFN I-related genes, and they exhibit dampened transcriptional priming of several components of the NLRP3-inflammasome (192). Consequently, bats show almost no inflammatory immune response to cytoplasmatic or circulating self-DNA (191–193).

Apoptosis represents a form of cell death of somatic cells long since known to be associated with abundant translocation of nuclear contents to cell surface blebs and ultimately, to the blood stream (195–197). Even more DAMPs are released when immune cells undergo inflammatory forms of cell death such as pyroptosis (22, 58, 181, 198, 199). The DAMP most frequently reported for COVID-19 is lactate dehydrogenase, a central enzyme of anaerobic glycolysis that is present in virtually all cells and that, therefore, reliably indicates cell death when occurring at high concentrations in the extracellular space (200–202). Other DAMPs which have been observed at elevated levels in the blood or plasma of patients with severe COVID-19 comprise high-mobility group box protein (HMGB)1 (51, 110, 111, 202–206), S100 proteins (108, 110, 202, 207–209), plasma hyaluronan (210), extracellular (e)ATP (211, 212), the antimicrobial peptide LL-37 (213, 214), histones (e.g (215–217), reviewed in (218, 219) and circulating self-DNA, including nDNA (13, 20, 21, 42, 46, 47), mtDNA (13, 18, 21, 42, 43, 48, 49), NET-associated cfDNA (201, 215, 220–222), histone-DNA complexes (219), and cfDNA of non-specified subcellular origin (45, 48, 50–54). As recently reviewed for sepsis (72), most of these studies reported a positive correlation of the plasma levels of at least some of the beforementioned DAMPs with the degree of disease severity (e.g., mild versus severe cases, COVID-19 patients at ICU admission versus healthy controls, ICU-admitted cases with fatal outcomes versus surviving patients, poor oxygenation status, patients with acute respiratory distress syndrome or with multisystem inflammatory syndrome in children, etc.).

As predicted by Polly Matzinger (62, 223), DAMPs trigger the activation of antigen-presenting cells, act as chemokines that recruit dendritic cells, macrophages, or T cells to the site of injury, or prime immune cells, i.e., they contribute to ‘trained immunity’ (58–60, 62, 224–226). The signalling function of lactate dehydrogenase remains a matter of discussion (227), but DAMPS sensu strictu such as HMGB1, eATP and cytoplasmatic and extracellular dsDNA prime the immune system for stronger responses to current and upcoming threats. One of the best studied examples of a DAMP-mediated immune priming is the activation of inflammasomes in two steps (59, 176, 177). Multiple lines of evidence support a critical role of inflammasomes in COVID-19 and – likely - its autoimmune complications (12, 13, 22, 29, 39, 181, 228–232) (for a review, see (39)). The lung and spleen tissue obtained from patients who died from COVID-19 exhibited higher densities of cells expressing NLRP3, IL-18, NF-κB and gasdermin D, and even HMGB-1, than age-matched controls who had died unexpectedly, but free of SARS-CoV-2-infection (13, 233). This observation indicates that the cells were already in the primed stage. Moreover, exosomes from patients with severe COVID-19 (but not light COVID-19 or healthy donors) induced the expression of NLRP3, pro-caspase-1 and pro-IL-1β in human endothelial cells, microvascular endothelial cells and liver endothelial cells (232). This last study appears to be the first empirical proof of concept showing that cell content released locally from infected cells can trigger systemic inflammatory effects in COVID-19. In addition, enhanced plasma levels of pyroptosis markers were detected in COVID-19 patients, and the levels of caspase -1 and IL-18 in serum correlated with the degree of COVID-19 severity (12, 22, 29, 220, 221), particularly in elderly patients (234). Neutrophils, macrophages and PBMCs from SARS-CoV-2-infected patients also exhibited active AIM2 or NLRP3 inflammasomes and enhanced expression of ASC-speck protein, caspase 1 or gasdermin-D (12, 13, 22, 29, 199, 221, 233). Correspondingly, specific inhibition of the NLRP3 inflammasome suppressed immune overactivation and alleviated COVID-19-like pathology in mice (235). Intriguingly, the SARS-CoV-2 N protein interacts directly with NLRP3 and promotes its binding to ASC, thereby facilitating the assembly of the NLRP3 inflammasome (236). Moreover, the E protein can form cation channels that allow for pyroptosis in the absence of active gasdermin-D (237).

“Unless cell death is explicitly assayed in an experiment, the contribution of dying cells to the generation of DAMPs and alarmins may be missed”. Melinda Magna and David S. Pisetsky, 2016 (61)

For many years, self-DNA was considered as immunologically inert, because self-DNA released from dying cells into the extracellular space was believed to be inaccessible to the intracellular DNA sensors. However, when bound to peptides or proteins, extracellular DNA can translocate into specialized compartments, including the endosome of pDCs, and then gain immunogenic properties (61, 72), e.g. via recognition by TLR9 (238–241). As pointed out earlier for the HIV-1-context (65) or sepsis (72), pyroptosis releases self-DNA together with cellular content, including HMGB and histones, i.e., two types of DAMPs which intrinsically possess DNA-binding properties. In fact, the nuclei of cells that die via pyroptosis usually remain intact and only show chromatin condensation, meaning that rather than as ‘naked’ DNA, genomic DNA is released as chromatin, hence, in an immunogenic form (177). Intratracheal instillation of mouse alveolar epithelia with the SARS-CoV-2 Spike protein and poly (I:C), a synthetic RNA mimic, triggered lung tissue injury and enhanced levels of circulating TNF-α and HMGB1 (203). The crucial role of HMGB1 in this effect is underlined by the observation that a treatment with anti-HMGB-1 antibody reduced these detrimental effects (203). Thus, a DAMP can increase the detrimental effects of virus-derived PAMPs, for which reason it seems likely that the positive association of diverse DAMPs with COVID-19 severity reflects a causal role, rather than being merely correlative. Also in NETs, self-DNA is bound to HMGB1 or histones, which prevents its degradation and facilitates its passage through membranes or its active uptake via endocytosis and/or receptor binding. In consequence, NETs are known to cause sterile inflammation or small vessel vasculitis, thereby generating microvascular damage that contributes to injury in the brain and lung tissues or favours certain autoimmune pathologies, including SLE (83, 242–244).

In addition, several host defence peptides (also termed ‘small secreted antimicrobial peptides’) possess DNA-binding properties (reviewed in (238)), and among these, LL-37 has been reported at increased levels in plasma of COVID-19 patients (213, 214). LL-­37 enhances the IL-­1β-­induced production of cytokines such as IL-6 in monocytes, likely via a DNA-independent mechanism (238), while extracellular histones induce the secretion of IL-1β, IL-6 and TNF-α from circulating monocytes (245). These cytokines, in turn, trigger a pro-inflammatory senescence-associated secretory phenotype in human umbilical vein endothelial cells, at least when they act in synergy (246). Thereby, cytokines that are released via pyroptosis can trigger the pyroptosis-independent release of cellular content from certain immune cells. Synergistic effects can also occur via the inflammasome-mediated axis: complexes formed by HMGB1 and dsDNA induce caspase-1 activation and IL-1β–release from monocytes at 25 times lower concentrations than each molecule alone, because they activate the AIM2 inflammasome (247).

A further, universal DAMP that is well known to act in synergy with other DAMPs or otherwise facilitates their perception is eATP. Extracellular ATP was found at elevated concentrations in the blood of patients with severe COVID-19 (211, 248). The increase in eATP concentrations could be related to two mechanisms: reduced activities of ectonucleotidases (211, 248) and opened pannexin channels (249). The eATP-mediated activation of P2X7 and the resulting production of intracellular reactive oxygen species (ROS) (250) can activate the NLRP3 inflammasome and the release of IL-1β, IL-12 and IFNγ from macrophages (251).

Anti-dsDNA autoantibodies represent a further class of proteins that enhance the accessibility of extracellular DNA to innate immune sensors. The release of cellular content during pyroptosis arguably means a massive liberation of potential autoantigens within an already pro-inflammatory environment. Therefore, multiple lines of evidence link inflammasome-mediated pyroptosis to the generation of anti-nuclear and/or anti-dsDNA antibodies. The detailed mechanisms that lead to the formation of these types of autoantibodies remain under discussion, although it seems reasonable to argue that the formation of antinuclear antibodies – a crucial pathogenic feature of SLE that has also been reported for patients following vaccination with BNT162b2 or mRNA-1273 (126) – is evidently facilitated by the fact that the nuclei of cells undergoing pyroptosis usually remain intact. More importantly in the context of the present considerations, DNA-containing immune complexes - like other DNA/protein complexes - present another means of facilitated uptake of cfDNA into immune cells, e.g., into monocytes (252), i.e., they allow for the sensing of self-DNA by TLR9 and thereby stimulate cytokine production (241, 252–254). In other words, pyroptosis releases self-DNA under conditions that are likely to render immunogenic properties to extracellular DNA, generating circulating self-DNA that is particularly prone to promote inflammation and autoimmunity (83, 244).

An enhanced prevalence of autoantibodies against nuclear self-antigens – including dsDNA - has been reported for patients with severe COVID-19 (86, 125, 255–258), in particular those patients who developed an autoimmune disease as a consequence of COVID-19 (259, 260). In addition, these autoantibodies were detected in the blood of patients who presented some of the adverse effects of the mRNA vaccines (261). For example, elevated levels of anti-nuclear and anti-dsDNA antibodies and, in one case, even anti-histone antibodies, occurred in various patients who developed autoimmune hepatitis after the first dose (132) or new onset SLE after the second or third dose of the BNT162b2 mRNA vaccine (129, 131, 132) or the mRNA-1273 vaccine (126).

At least one PRR is capable of sensing dsDNA outside of the cell: Although RAGE had been discovered as a receptor of advanced glycation end products, it functions as a multi-ligand PRR that senses diverse DAMPs, including HMGB1, S100 proteins and dsDNA, to activate NF-κB. RAGE is a transmembrane protein expressed in endothelial cells, pneumocytes, T and B cells, alveolar macrophages, monocytes and dendritic cells (262). In addition, proteolytic cleavage of the extracellular portion of membrane-bound RAGE can release soluble forms of RAGE, a process which is upregulated by inflammatory signals. Increased soluble RAGE in serum is also a predictor of mortality among COVID−19 patients (31, 32). The expression of RAGE was significantly increased in patients with severe COVID-19, along with its ligands, including S100 and HMGB-1 (263, 264), and a system-wide transcriptomic analysis identified RAGE among the strongly upregulated genes in the liver, and among the slightly, but significantly upregulated genes in the heart and lymph nodes of COVID-19 patients (Supplementary Table S2 to (265)). The causal involvement of RAGE in vascular injury and severe disease in COVID-19 patients was underlined by the reduced systemic inflammation and damage to blood vessels and increased survival of mice treated with pharmacological inhibitors of RAGE (262). Of particular interest in the context of the present work, RAGE can trigger inflammation by activating NF-κB, and it can sequester extracellular DNA to facilitate its transport to the endosome and thus, its exposure to TLR9 (30). Moreover, RAGE can trigger a prolonged activation of NF-κB that apparently overcomes several endogenous negative feedback control mechanisms (reviewed in (24)). Since RAGE itself is a NF-κB controlled gene, RAGE is particularly prone to contribute to potentially fatal feedforward scenarios, and NF-κB has been suggested to ‘universally enhance STING-mediated immune responses’ (266).

The SARS-CoV-2 Spike protein, CoV-2 S, is the target encoded by the mRNA vaccines (267, 268). Although mRNA is normally assumed to have a short persistence time, it is highly likely that CoV-2 S is expressed on the cell surfaces of vaccinees over a considerable timespan (2, 17, 36, 182, 186, 269, 270). Vaccine mRNA has been detected at least 15 days after the first or second dose of BNT162b2 (271). Since only one (negative) sample covered a later time point (27 days) in this study, it seems likely that the persistence time of vaccine mRNA is longer than 15 days. In fact, another group detected vaccine mRNA in blood samples from 10 of 108 Hepatitis C Virus-infected patients at 28 days after vaccination (272), and in the germinal centres of axillary lymph nodes, vaccine mRNA was even detected at significant abundances at 37 days post vaccination and remained detectable at 60 days post vaccination (273). Similarly, spike itself was detected in the endothelial cells within inflamed areas of the brain and heart of a man who died three weeks after receiving his third vaccine dose with BNT162b2 mRNA (274), and in the plasma of individuals who exhibited postvaccination myocarditis, most of them within a week after vaccination with the BNT162b2 or the mRNA-1273 vaccine (134). In another study, circulating S1 was detected in the plasma of eleven out of 13 participants and, although it peaked on average 5 days after receiving the first dose of mRNA-1273, Spike was detected still after 15 days (275). Plasma S protein levels of 10 ng ml^-1 were observed 10 days after vaccination in a woman with mRNA-1273-induced thrombocytopenia, while S1 concentrations reported for plasma of COVID-19 patients oscillate around 50 pg ml^-1 and can reach maximum levels about 1 ng ml^-1 (reviewed in (276)). Considering in addition that Spike expression in vaccinees is likely to occur in tissues and organs that in SARS-CoV-2-infected individuals are unlikely to be reached by circulating CoV-2 S, potential effects of Spike on host immunity and other vital functions must be taken seriously.

In 2020, Jiang Hui and Mei Ya-Fang presented the hypothesis that CoV–2 S impairs DNA damage repair and thereby reduces the efficiency of an essential step in antibody production: the adaptive production of diverse antigen receptors via regulated dsDNA breaks and their subsequent repair (277). This work has been heavily criticised and finally was retracted. One of the critical voices was Derek Lowe (278) who, among other points, highlights that Jiang & Mei transfected DNA into specific and cultured cell lines, which do not necessarily respond like cells in vivo. Lowe reassures that - although “a lot of people are worried about Spike protein circulating around through the body”, “there is no evidence (and no particular reason to believe) that circulating Spike protein after vaccination, such as it is, gets taken up into other cell types and then taken into their nuclei” (278). Indeed, I am not aware of a report on CoV-2 S entering the nucleus, and cytosolic S protein does likely not trigger inflammation in epithelial cells (279). Still, I think that there is enough evidence that (and how) CoV-2 S can damage cells and – in particular – their DNA without entering the nucleus.

Evidence for a DNA-damaging effect of SARS-CoV-2 - and perhaps CoV-2 S - is accumulating (23, 36, 280–283). Comet assays and γ-H2AX immunostaining revealed elevated levels of DNA damage in SARS-CoV-2-infected Huh7 or Vero-6 (African green monkey kidney) cells (66, 190, 284, 285), but also in lymphocytes and cardiac tissue of deceased COVID-19 patients (286–288). In the human studies, DNA damage levels correlated with Il-6 expression in infected cells (285) and serum levels of IL-6 (287) and of other inflammatory ILs (286). Post-mortem transcriptomic analyses of cardiac tissues of COVID-19 patients revealed an enrichment of DNA damage and repair, heat shock, and cell cycle control among the predominant upregulated genes (288). An increase in oxidative stress and in the number of DSBs has been observed in PBMCs from older individuals at 24 h after vaccination with BNT162B2 (280), and several p53-controlled genes, including those related to apoptosis and DNA-repair, were overexpressed in PBMCs from a patient who developed myocarditis after BNT162b2 vaccination (hence, likely as a consequence of CoV-2 S expression in the absence of infection) (289). The consequences of these alterations comprise cell cycle arrest in the S1-phase, the activation of a pro-inflammatory senescence-associated secretory phenotype that exhibits elevated resistance to programmed cell death (190, 246), and enhanced expression of NF-κB and of the ACE2 promoter (290, 291).

In fact – as described earlier for HIV-1 (15, 292–294) - SARS-CoV-2 induces NF-κB, which inhibits the DNA damage-activated transcription factor p53. P53 is involved in cell-cycle control as part of the DDR, including the decision ‘DNA repair versus elimination of cells’ (295, 296). Correspondingly, p53-controlled genes were overexpressed in leukocytes from patients with severe COVID-19 (297). Moreover, SARS-CoV-2 degrades checkpoint kinase (CHK)1 (190), an effector downstream to Ataxia telangiectasia and Rad3 related protein (ATR) that earlier had been demonstrated to be induced in SARS-CoV-2-infected Vero-6 cells (284). Consequently, an inhibitor of ATR blocks the replication of SARS-CoV-2 after entering cells and thus, was identified as a potential anti-COVID drug that exhibits antiviral activity against SARS-CoV-2 in diverse cell types (298).

Zhou et al. (17) were among the first who discovered that the switch from a suppression to an (over-)induction of cytokine signalling associates with syncytia formation. Intriguingly, they also reported that CoV-2 S expression is sufficient to fuse cells. Several independent studies confirmed syncytia-formation by virus-free, Spike-expressing cells (23, 186, 269, 299–302). Syncytia formation, in turn, is inevitably associated with micronuclei formation, nuclear membrane blebbing and therefore, DNA damage, which ultimately leads to cell death (17, 23, 36). Later work confirmed that syncytia formation triggers IFN I signalling (23).

If we swich from the human perspective to the perspective of the virus, CoV-2 S-mediated detrimental effects of SARS-CoV-2 on host DNA become something that one simply would expect. Common knowledge holds that coronavirus Spike proteins control binding to the ACE2 receptor and subsequent fusion of the viral and the plasma - or endosomal membrane - of host cells, to facilitate viral entry (2, 186, 270) ( Figure 3 ). However, viruses that replicate in the cytosol are short of raw material, i.e., nucleic acids, viruses gain a fitness advantage from supressing or escaping from the host immune system, and all viruses are under selective pressure to make the most efficient use of their small set of genes. Therefore, being a protein evolved to facilitate membrane fusion and the formation of pores, the Spike protein of SARS-CoV-2 has at least two additional jobs: it is employed to destabilise the nuclear envelope, and for syncytia formation (33, 34): a process associated with DNA damage, including the exposure of chromatin to cytoplasmatic DNA sensors (34, 36). For the latter, CoV-2 S possesses a signal that facilitates trafficking to the cell surface (269) and therefore is also expressed on the membrane of infected cells (182, 186, 269). Both effects benefit the virus: while nuclear membrane blebbing facilitates the liberation of the required ‘raw material’ from the nucleus, syncytia allow for viral spread without exposure to host antibodies, NETs or macrophages ( Figure 3 ).

Evidently, more research will be required in this context. Most of the studies that reported syncytia-formation in virus-free systems that employed widely used model cell lines like HEK293T (human kidney cells), HeLa (human cancer cells), Calu-3 (human lung cells) or Huh-7 (human liver cells), which are overexpressing ACE2, and the used plasmid-based transfection systems further contribute to generate artificial conditions that limit a direct translation of the obtained results to realistic in vivo situations. On the other hand, the Spike protein encoded by the mRNA vaccines is stabilised in the prefusion-conformation to facilitate ACE2-binding and cell entry (303). It would be interesting to investigate to which degree syncytia formation is affected by this change versus ACE2 expression level (17, 23, 36, 269). In short, the virus uses CoV-2 S to damage it´s host´s DNA, and thereby enhances the lifetime of the infected cells and at the same time, upregulates ACE2 expression to decorate more cells with its entry receptor. By doing so, CoV-2 S activates certain elements of innate immunity which are also induced downstream to dsDNA sensing. Since CoV-2 S decorates more cells with the SARS-CoV-2-entry receptor, expression of this protein alone can be sufficient to trigger cell-to-cell fusion and subsequent nucleus-to cytosol shuttling of chromatin (17, 23).

Therefore, a scenario in which vaccination leads to an overstimulation of the innate immune responses downstream to dsDNA-sensors that generate the mentioned adverse effects appears definitively possible. In the following, I compile information on how CoV-2 S and SARS-CoV-2 infection affect elements of the dsDNA-sensing machinery and then develop a worse-case scenario describing hypothetical effects of Spike expression in an innate immune-environment that has already been primed by a prior SARS-CoV-2 infection. To base this scenario on – admittedly indirect – empirical evidence, I first highlight similarities between SARS-CoV-2 and HIV-1 that are related to HIV-1’s effects on DNA damage, dsDNA-sensing, inflammation, and autoimmunity.

“Although there is much to learn about SARS, many lessons can already be drawn from our experience with HIV”. Anne S. De Groot, 2003 (93)

Interfering with the DDR or directly damaging their host’s DNA, syncytia formation, and the induction of bystander cell death, as drivers of inflammatory responses that harm the host and benefit the virus: all these particularly important features are shared among HIV-1 and SARS-CoV-2 (304–306). Several inflammatory and autoimmune-related pathologies coincide among HIV-1-infected individuals, COVID-19 patients and vaccinated individuals who suffer from severe adverse effects. Based on the knowledge available to far, it seems very likely that these similarities result from the same mechanisms acting in all three contexts. Likewise, an enhanced prevalence of autoantibodies against nuclear self-antigens had already been reported for patients with HIV-1, particularly for HIV-1-infected patients presenting thrombocytopenia (91, 98, 307, 308). The present work is far from being the first one to highlight HIV-1 as an example from which a lot can (or could have been) learned for faster and more efficient response to the COVID-19 pandemic (93, 94, 309), and even the global phosphorylation landscape of SARS-CoV-2 infection revealed extensive similarities with the patterns in protein activation early during HIV-1 infection (310).

Since the early nineties of the last century, HIV-1 is known to damage the DNA of its host and activate genes involved in the DNA damage response (DDR) (294, 311–314), and HIV-1 is also well known for its ability to fuse its host cells and thereby reach 4-5 times higher reproduction rates (304, 315). In the case of HIV-I, the bipartite envelope glycoprotein (Env) performs the two essential functions of binding to receptors on the surface of target cells and fusion among host-cell and viral membranes, including the formation of a fusion pore to deliver the viral core into the cell cytoplasm (306). Apparently, the virus employs the same protein to destabilize the nuclear envelope. In this context, it seems important to recall that SARS-CoV and SARS-CoV-2 are closely related and that consequently, the S2 domain of the SARS-CoV spike protein and of CoV-2 S are highly similar (2, 34, 270). Importantly, SARS-CoV spike, in turn, shares multiple structural similarities with the gp41 unit of the envelope glycoprotein of HIV-1 (135, 316). Based on these similarities, which are underlined by the cross-reaction of CoV-2 S-directed non-neutralizing polyclonal antibodies with gp41 (317), these proteins can be expected to interact with cell membranes via the same mechanism.

Furthermore, while there should be no need to repeat that HIV-1 infects mobile immune cells (in particular CD4+ T helper cells), it seems worth to mention that evidence for SARS-CoV-2 doing the same is accumulating. SARS-CoV-2 has been shown to infect monocytes, macrophages and B-cells (22, 28, 318), and at least Pontelli et al (318) present evidence supporting a successful reproduction of the virus in these cells, although at low rates. Since excessive infiltration of mobile pro-inflammatory cells such as macrophages and T-helper 17 cells has been found in lung tissues of patients with COVID-19 (11), it appears at least possible that SARS-CoV-2 can apply a strategy that was considered unique to HIV-1: triggering local inflammation to attract mobile immune cells and infect these cells to achieve systemic distribution throughout the host.

A further shared element is the importance of eATP as a DAMP: HIV-1-infected target cells release ATP, which then acts on purinergic receptors to stimulate fusion between Env-expressing and CD4+ expressing membranes (319). eATP also favoured the infection of microglia with HIV-1, an effect that was associated with elevated levels of IL-6 and IL-18 and with changes in p53 activity (320). The relevance of this is underlined by the observation that an inhibitor of P2X receptors effectively inhibited cell-to-cell transfer of HIV-1 from productively infected CD4+ lymphocytes (321). Therefore, therapeutic applications of P2X7R antagonists seem promising tools to control infection with HIV-1 (322), but also SARS-CoV-2, but also SARS-CoV-2 ( Table 1 ).

Although viruses usually supress or avoid host immunity, CoV-2 S has been suggested to induce inflammation via a TLR2 (or TLR4) – and MyD88-dependent dependent activation of the NF-κB pathway in human and mouse macrophages (279), and also to trigger lung cancer progression, again via TLR2 (331). Likewise, intravenous administration of Spike or a stimulation of cultured cells with CoV-2 S induced the expression and release of TNF-α, IL-1β, IL-6 and IL-18 in PBMCs, macrophages, monocytes, lung epithelial cells, human umbilical vein cells as well as in lung, liver, kidney, and eye tissues (279, 332, 333). In fact, the S1 subunit alone was observed to induce an increased production of IL-6 and activation of NF-κB and in consequence, inflammation in endothelial cells (334), and circulating S1 induced the expression of TLR2, TLR4, NLRP3, IL-1β, TNFα and HMGB1 in rats at 24 h after treatment, and TLR2, TLR4, NLRP3 and IL-1β remained overexpressed even after seven days (335). Spike also has been reported to activate caspase 1 and the NLRP3 inflammasome in hematopoietic stem/progenitor cells and endothelial progenitor cells (336). Intriguingly, exposure to CoV-2 S of human umbilical vein cells activated NF-κB and ACE2 (333), while vice-versa, administration of an ACE2 inhibitor blocked the activation of inflammasome components by Spike (336). Quantification of lactate dehydrogenase (LDH) enzymatic activity in the culture medium confirmed pyroptotic cell death of these cells (336). Most of the beforementioned studies used commercially available, recombinant spike proteins produced in human cells (279, 333, 335), hamster cells (332), or E. coli (334), which opens the possibility of contaminations with immunogenic molecules that originate from the expression system. For example, bacterial lipopolysaccharides at very low concentrations as they have been detected as contaminations in commercial proteins activated human immune cells (337), and these endotoxins also signal via TLR4 (338). However, at least Sung et al. (339) used a pseudotyped lentivirus carrying the SARS-CoV-2 Spike protein, hence, a situation that resembles the vaccines. Assuming the reported in vitro-effects of CoV-2 resemble the situation in vivo, we can expect them to apply also to an infection with SARS-CoV-2. Indeed, nasopharyngeal epithelial cells of COVID-19 patients exhibited significantly higher expression of TLR2 and TLR4 as compared to controls (340), and TLR4 and its downstream elements (including Myd88, IRAK1 and TRAF6, and NF-κB - dependent genes) were significantly upregulated in PBMCs from 20 human COVID-19 patients (341).

The studies cited in 5.4. demonstrate that infection with SARS-CoV-2 S can prime the innate immune system for faster and stronger responses to any subsequent infection or other cell-damaging event. Indeed, plasma from COVID-19 patients exhibited increased in P2X7 expression (342). eATP can act as a second signal that activates inflammasomes, as shown by the observation that exposure to CoV-2 S followed by eATP triggered a stronger expression of pro-IL-1β, ASC, NLRP3 and gasdermin D in macrophages derived from COVID-19 patients than in SARS-CoV-2 naïve cells, and only patient-derived macrophages exhibited active ASC specks and increased secretion of TNF-α and IL-1β (229). Importantly, altered inflammasome and stress responses persisted after short-term patient recovery (29) and the differential responsiveness was maintained even by macrophages from fully convalescent COVID-19 patients after more than 50 days (i.e., after several cycles of monocyte renewal) (229). Likewise, TLR4 and its downstream elements (including Myd88, IRAK1 and TRAF6, and NF-κB - dependent genes) were significantly upregulated in PBMCs from 20 human COVID-19 patients (341).

These studies confirm that infection with SARS-CoV-2 causes a long-term reprogramming of the immune system, particularly in macrophages. Thus, it seems plausible that in vaccinees who had an infection before being vaccinated, the effects of SARS-CoV-2 expression (including DNA damage) occurred in the context of a primed DNA-sensing machinery: a situation that can strongly enhance its immunogenic potential. Others discovered that S1 and S2 proteins administered intraperitoneally triggered enhanced concentrations of IL-6, IL-1β, and TNFα (16 hr post treatment) in WT mice but not in mice lacking TLR2, which indicates a role of TLR2, rather than TLR4 (279). In addition, CoV-2 S induced an enhanced release of ATP and IL-1β from human lung epithelial cells (249) and cultured microglial cells (BV2 line) and in the latter, it also induced the expression of the purinergic eATP receptor P2X7 (343).

I conclude that SARS-CoV-2-mediated immune priming can enhance the DNA-damaging and pro-inflammatory effects of CoV-2 S and cause certain responses to pass a threshold or point of no return, reaching those dimensions that we see in the severe adverse effects of the Spike-based mRNA vaccines. Even phase III vaccine trials usually exclude individuals who show preexisting immunity due to previous infection, but the vaccination campaigns included significant proportions of the entire population at a time point at which many people had already passed through an infection with SARS-CoV-2, and pre-existing immunity was seldom checked in these mass vaccination events. In summary, immune priming represents an example of a mechanism that could generate different outcomes of vaccination depending on an earlier – perhaps non-symptomatic and not detected – infection with SARS-CoV-2.

Most of the published work on the SARS-CoV S protein and the similarities of SARS-CoV-2 with HIV-1 was ignored in the searches for new or pre-existing drugs to treat COVID-19 patients, and in the scientific activities involved in vaccine development. Perhaps as a consequence, on the one hand, patients were treated with IFN and indeed, slightly (although not significantly) more people died in the treatment group (344). On the other hand, DNases or DNA scavengers were hardly considered: to the best of my knowledge, all studies in which patients were treated with Dornase α are cited in chapter 3.5 (116–119, 121, 188, 222), and the obtained results – in particular the survival of all involved patients - clearly demonstrate the beneficial effects of anti-DNA treatments. As suggested by Anne de Groot and an international group with lead author Nevan Krogan (93, 310), existing knowledge should be considered even when it must be found in 20 years ‘old’ papers. Moreover, searches for similar pathogens should define ‘similarity’ with a focus on shared strategies or functions of the pathogen that generate similar effects on the interactions with the host, rather than on ‘taxonomic similarity’ defined at the level of sequences.

“We did not expect an RNA virus like SARS-CoV-2 to be sensed by the DNA sensor AIM2”. Caroline Junqueira et al., 2021 (345)

In addition to the beforementioned issues, it seems that the dominance of the Janeway paradigm (346, 347) has significantly hindered a full appreciation of the immunogenic and pro-inflammatory effects of self-DNA. In fact, the author is aware of a few concrete cases in which the publication of results supporting the immunogenic effects of self-DNA or the danger model in general was significantly slowed down (personal communications by Polly Matzinger and Verena Kopfnagel, and own observations (348)). As a consequence, it seems that cfDNA, and nDNA in particular, remains strongly under-investigated. Even among those studies that originally linked the inflammation-related complications in COVID-19 to dsDNA sensors, only a single one has directly quantified mtDNA and nDNA and none has treated cells with natural DNA, as done, for example, in some of the studies that investigated the role of cfDNA in trauma (69). Nevertheless, even among the classical studies on trauma, three studies focused on mtDNA (66, 67, 170), although the fourth study found that serum IL-6 levels, inflammation and critical illness correlated with the levels of nDNA, not mtDNA (68). Likewise, most of the clinical studies that quantified plasma DNA concentrations as a possible marker of severe COVID-19 have quantified only mtDNA, without citing any empirical evidence for the assumed non-activity of nDNA.

Evidently, future work will have to provide definitive evidence for a causal role of cfDNA in severe COVID-19 or the vaccine-triggered adverse events. Still, beyond doubt, cfDNA levels in the blood or plasma are a useful biomarker of disease severity in COVID-19 patients and likely, a predictor of certain adverse effects of the vaccines. Although it would be difficult to follow the advice of The European Academy of Allergy and Clinical Immunology (EAACI) that “all clinical and laboratory information should be collected and reported…to understand the specific nature of the reported severe allergic reactions” (349), it seems mandatory to include cfDNA of both mitochondrial and nuclear origin and anti-dsDNA antibodies in the list of standard laboratory information that should be obtained for patients that suffer from autoimmune-related or chronic inflammatory pathologies and for all participants in future vaccination trials.

“…studies … are often performed using a single well-defined ligand. However, …cells usually receive multiple inputs or experience many environmental alterations simultaneously.” Andrea Ablasser, 2019 (123)

Multiple reports on stronger responses to a certain trigger shown by cells from COVID-19 patients or in response to DNA bound to other DAMPs provide examples of clinically important synergistic effects. In particular, extracellular self-DNA gains immunogenic properties when bound to other DAMPs or extracellular vesicles, independently of the concrete pathosystem. Therefore, the suggested effects can occur in basically all situations that comprise tissue damage, particularly when this damage occurs in the context of pre-activated innate immune signalling. The antibody-assisted infection of cells and the priming of certain immune responses by viral proteins represent further examples of synergies that can create a significant level of context dependency in the obtained results. As outlined by Andrea Ablasser, such effects are likely to be overlooked in experiments that expose naïve cell lines under controlled conditions to a single ligand (123). If we, e.g., study human defensin 5 (HD5) and the human cathelicidin known as LL-37 each in an isolated manner, we will observe that LL-37 binds to the carboxypeptidase domain of human ACE2 even stronger than HD5. From this observation, we could conclude that LL-37 bears a great potential to be tested as an anti-SARS-CoV-2 peptide, because it blocks the entry receptor of the virus (350). However, in vivo, the release of LL-37 and human β-defensin 3 might take place in the presence of pyroptotic cells or other sources of cfDNA. In case that – as shown earlier – these proteins facilitate the uptake of self cfDNA into pDCs and monocytes (239) and thereby potentiate its immunogenic potential to trigger pro-inflammatory effects through the TLR9 pathway (351), the net outcome of treating severe COVID-19 patients at the later stages of the disease with LL-37 might be fatal, as SARS-CoV-2 infection has significantly declined, while the inflammatory response escalates dramatically, becoming predominant (67).