Pathogenic antibodies induced by spike proteins of COVID-19 and SARS-CoV viruses

This study, using a virus-free mouse model, explores the pathogenic roles and novel mechanism of action of certain antibodies specific to the spike proteins of highly pathogenic coronaviruses such as the COVID-19 and the SARS-CoV viruses. These pathogenic antibodies, induced during a highly pathogenic infection such as the COVID-19 infection, target and bind to host vulnerable cells or tissues such as damaged lung epithelium cells, initiate a persistent self-attack immune response, and lead to serious conditions including ARDS, cytokine storms, and death. Moreover, the pathogenic antibodies may also be responsible for infection-related autoimmune diseases, including those experienced by COVID-19 long haulers. Furthermore, the pathogenic antibodies can bind to the unmatured fetal tissues and cause abortions, postpartum labors, still births, and neonatal deaths of pregnant females. Novel clinical interventions, through disrupting the binding of these pathogenic antibodies, can be developed to fight the COVID-19 pandemic. In addition, the new concept explored by this study may be applicable to other infectious diseases, such as the highly pathogenic influenza infections.


1
The pandemic of coronavirus disease 2019 (COVID-19) is a major threat to worldwide population health and economies 1,2 . One week after onset of the COVID-19 infection, the clinical condition of the disease can become severe, progressing with hypoxemia and dyspnea, and rapidly develop to acute respiratory distress syndrome (ARDS) in 17% of patients. 65% of these patients worsened and died due to multi-organ dysfunction 3,4,5 . Similar clinical characteristics have also been observed in patients infected with other highly pathogenic respiratory viruses such as the severe acute respiratory syndrome (SARS) virus (SARS-CoV) 6 , the middle east respiratory syndrome (MERS) virus (MERS-CoV) 7 , and the avian H1N5 and H7N9 influenza viruses 8,9 . In some cases the viral load was low or undetectable at the time of severe illness 10 , suggesting that there were other causes of death than the virus alone. It has been reported that over-reacting immune responses as well as the accompanying cytokine storms played a critical role in the deaths of the infected 11,12,13 , yet what initiated the cytokine storms remains a mystery. We have made the unexpected discovery in a mouse model that injection of anti-influenza sera into pregnant mice induced lung inflammation in mouse pups born to the dames 14 . Thus, we suspected that certain antibodies induced by a highly pathogenic virus may be pathogenic themselves, through the targeting of host cells or tissues. To prove this hypothesis, we investigated the pathogenic effects of anti-coronavirus antibodies, including anti-COVID-19 viral antibodies, in this study. virus were tested for their binding to the human lung epithelium cell line A549. Seven of the antibodies tested were monoclonal antibodies specific to the SARS-CoV-2 spike protein. These antibodies have been previously reported by others and were isolated from the B cells of COVID-19-infected patients 15,16,17,18,19,20 . These seven monoclonal antibodies were reproduced according to the gene sequences encoding the antibodies, from the NCBI crystal structure site ( Table  1). The other anti-coronavirus antibodies tested include polyclonal rabbit anti-SARS-CoV-2 spike one (anti-COVID-19 S1), rabbit anti-SARS-CoV spike glycoprotein (anti-SARS S), and monoclonal mouse anti-SARS-CoV nucleocapsid (N) proteins (anti-SARS N). These antibodies were commercially available (Bioss Antibodies) and were produced by immunization of animals with related recombinant proteins of the SARS-CoV-2 and SARS-CoV viruses. The polyclonal anti-SARS-CoV-2 nucleocapsid (N) protein (anti-COVID-19 N) antibodies were obtained by immunization of rabbits with the recombinant nucleocapsid (N) protein of SARS-CoV-2 virus (HuaAn McAb Biotechnology). Binding of those antibodies to healthy (intact) or damaged lung epithelium cells were tested with A549 cells. In order to induce damaged cells, A549 cells were  21 . The fluorescent labeled wheat germ agglutinin (WGA), which specifically binds to N-Acetylneuraminic acid (Vector), and a flow cytometry analysis were used to determine the levels of sialic acid on the surface of A549 cells. The damaged cells with missing sialic acid on the cell surface were used to imitate the in vivo conditions of infected lung epithelium cells (sick cells). As shown in Figure 1A, the level of sialic acid on the surface of A549 cells treated with sialidase was lower than those of untreated cells ( Figure 1A). The seven human monoclonal antibodies specific to the COVID-19 spike protein (Table 1) were tested for their binding to A549 cells. As shown in Figure 1C, two out of seven (28.6%) such antibodies, REGN10987 17 and B38 15 , strongly bound to the damaged A549 cells with missed sialic acid. REGN10987 also weakly bound to healthy A549 cells while B38 did not bind to the healthy A549 cells ( Figure 1B). The control antibody of CR3022-b6 20 did not bind to the healthy A549 cells nor the damaged cells ( Figure 1B and 1C). The other four human monoclonal antibodies bound neither significantly to healthy A549 cells nor to damaged cells. Further, the antibodies specific to the spike glycoprotein of SARS-CoV virus (anti-SARS S) strongly bound to the damaged A549 cells missing sialic acid, but not to the healthy A549 cells with sialic acid ( Figure 1E). In addition, the polyclonal antibody specific to the nucleocapsid (N) protein of COVID-19 virus (anti-COVID-19 N) and the antibody specific to the nucleocapsid protein of SARS-CoV virus (anti-SARS N) bound neither significantly to the healthy A549 cells nor to the damaged cells ( Figure 1D and 1E). Similar results were also observed with kidney embryonic cells of HEK-293 (data not shown). Taken together, the results of the in vitro assay indicated that certain antibodies specific to the spike proteins of the COVID-19 and SARS-CoV viruses have the potential to mislead the immune system to attack the host by binding to sick cells such as human lung epithelium cells in vivo. The antibody of REGN10987 may a higher potential of activating immune responses in vivo since it binds not only to sick cells but also to healthy cells, albeit at a lower rate.

Anti-COVID-19 spike antibodies alone induced systemic inflammation and injury in vivo
For further confirmation, a timed-pregnant mouse model without viral infection was used since the surface glycoprotein of fetal organs is not completely developed  or unmatured. The purified and endotoxin-free IgG of the anti-COVID-19 S1, anti-COVID-19 N, anti-SARS S, anti-SARS N, human monoclonal antibody of REGN10987 and B38 (Table 1) were used in the mouse model. The purified and endotoxin-free IgG of healthy sera from rabbit, mouse, and human, as well as the monoclonal antibody of CR3022-b6 were used as controls. Two dosages of each antibody IgG were injected intraperitoneally (IP) into timed-pregnant mice twice every three days at pregnancy (embryonic) days E15 and E18, respectively, as described in the methods (Figure 2A). The frequencies of sickness and death of the fetus and newborn mouse pups are summarized in Table 2 and Figure 2C. Injection of REGN10987 into pregnant mice induced significant fetal death and neonatal death of the mouse pups delivered to those dames ( Table 2). The fetal death was confirmed by autopsy ( Figure 2B). The results with this animal model indicated that REGN10987 has the highest potential for inducing sickness and death (61.9%), followed by B38 (45.8%) and the polyclonal anti-COVID-19 S1 (45.5%). The polyclonal anti-SARS S also caused significant sickness and death in the newborn mouse pups (37.6%). In addition, hyperemia at the end of left up and down limbs and a small hemangioma at the side of left eye of one pup was observed. The pup was born to a dame injected with the polyclonal anti-COVID-19 S1 antibody. Neither the healthy control antibodies nor the anti-COVID-19 N nor the anti-SARS N caused significant sickness or death of the newborn mouse pups ( Table 2). It was surprising that when the pathogenic polyclonal anti-COVID-19 S1 antibody was mixed with an equal amount of the non-pathogenic anti-COVID-19 N (50 g + 50 g), the sickness and death rate induced by the antibody mixture was significantly lower than that induced by the anti-COVID-19 S1 alone (Table 2). Moreover, the sickness and death rate induced by the highly pathogenic REGN10987 antibody also decreased significantly when a mixture of the antibody and other two non-pathogenic antibodies, CR3022-b6 and CC12.3, were injected ( Table 2). The mixture consisted of 40 g of REGN10987, 20 g of CR3022-b6, and 20 g of CC12.3. The data suggested that co-existing of non-pathogenic antibodies can reduce the pathogenicity of pathogenic antibodies. Histology changes. The tissue sections of lungs, brains, hearts, kidneys, intestines, and livers from the newborn Lung inflammation and injury. Acute lung inflammation was observed with the HE stained tissue sections from the mouse pups born to the dames injected with anti-COVID-19 S1, anti-SARS S, REGN10987, and B38 ( Figure 3). The lung lesion included pulmonary congestion, alveolar epithelial hyperplasia and thickening, hemorrhage, alveolar atresia, alveolar dilatation, and alveolar fusion ( Figure 3 and Figure S1). Infiltration of inflammatory cells at the local lesion areas were also observed. There were insignificant or minor histological changes with the lungs from the pups born to the dames injected with the antibodies of anti-COVID-19 N, anti-SARS N, CR3022-b6, and the control IgGs of human, rabbit, and mouse ( Figure 3).

Other organ inflammation and injury.
Inflammatory reactions and hemorrhage were also observed with the tissues of kidneys, brains, and hearts from the mouse pups, as mentioned above. The histology of the kidneys from the mouse pups delivered to the dames with the injection of anti-COVID-19 S1, anti-SARS S, B38, and REGN10987 showed acute tubular injury. Renal tubular epithelial cells showed granular or vacuolar degeneration, dilated or obstructed lumen, renal interstitial edema with a small amount of inflammatory cells infiltration, and some of the epithelial cells fell off (Figures 3 and S2). The kidney injury caused by REGN10987 was the most significant ( Figure 3). Furthermore, small amounts of cerebral hemorrhage or inflammatory cell infiltration was observed in the brains of mouse pups delivered to a dame injected with antibodies of anti-COVID-19 S1, anti-SARS S, B38, and REGN10987 ( Figure 3 and S2). Additionally, myocardial hemorrhage was observed in the hearts of the mouse pups delivered to the dames injected with anti-COVID-19 S1, anti-SARS S, and B38 ( Figure 3 and S2). Lastly, myocardial swelling and inflammatory cell infiltration were observed in a mouse pup delivered to a dame injected with antibodies of B38 ( Figure 3). There were insignificant or minor histological changes with the tissues from the pups born to the dames

Control MAb Human IgG
MAb-B38 MAb-REGN10987 REGN10987+2 MAbs Vehicle injected with the antibodies of anti-COVID-19 N, anti-SARS N, CR3022-b6, and the control IgGs of human, rabbit, and mouse. Taken together, the in vivo results of the virus-free animal model proved that certain antibodies specific to the spike proteins of the coronaviruses such as COVID-19 and SARS-CoV viruses can induce significant fetal and neonatal deaths and the systemic inflammation or injury of lung and other organs in vivo.
The results are consistent with the clinical observations of COVID-19 patients with severe illnesses 5 . Therefore, antibodies of anti-COVID-19 S1, anti-SARS-CoV S, B38, and REGN10987 are pathogenic and probably responsible for the serious conditions of a severe COVID-19 infection. On the other hand, the antibodies of anti-COVID-19 N, anti-SARS-CoV N, and CR3022-b6 are non-pathogenic since they did not induce significant adverse reactions in vivo.

Inflammatory cytokines
As further evidence of the pathogenicity of the antispike antibodies, the cytokine levels of monocyte chemotactic protein 1 (MCP-1), tumor necrosis factor- (TNF-), interleukin-4 (IL-4), IL-6, and IL-10 in the sera of mouse pups were tested by a multiplex Luminex assay kit (Millipore) according to manufacturer's instructions. The results of MCP-1 are summarized in Figure 4.
Both the pathogenic anti-COVID-19 S1 and REGN10987 induced significantly higher levels of MCP-1 ( Figure 4). Consistent with the surprising results of the histological changes, the treatment using the antibody mixture comprised of the pathogenic anti-COVID-19 S1 and the non-pathogenic anti-COVID-19 N, significantly reduced the cytokine levels of MCP-1 (P < 0.001) compared to that induced by the anti-COVID-19 S1 alone ( Figure 4). In addition, the treatment with the antibody mixture comprised of the pathogenic REGN10987 and two non-pathogenic antibodies of CR3022-b6 and CC12.3 also significantly reduced the level of MCP-1 (P < 0.001) compared to that induced by REGN10987 alone (Figure 4). The levels of other cytokines were not significantly elevated, probably due to the undeveloped immunity of the newborn mouse pups. The results were consistent with the results of the sickness and death rates ( Table 2) and the histology changes ( Figure 3). The data demonstrated that 1) pathogenic antibodies alone can induce high levels of inflammatory cytokines and have the potential to induce a cytokine storm or cytokine release syndrome (CRS); and 2) co-existence of non-pathogenic antibodies can reduce the inflammatory cytokine release induced by pathogenic antibodies and prevent the possible cytokine storm or CRS caused by the pathogenic antibodies.

In vivo antibody binding to multiple organs of mouse pups
As more evidence of the pathogenicity of the pathogenic anti-COVID-19 spike antibodies, in vivo antibody binding to tissues of mouse pups was detected using an immunofluorescent staining as described in methods. The human and rabbit anti-COVID-19 spike antibodies were significantly detectable at the inflammatory and lesion areas of the tissues of lungs, kidneys, brains, hearts, livers, and intestines from the mouse pups with severe sickness ( Figure 5). Those mouse pups were delivered to the dames injected with the pathogenic antibodies of anti-COVID-19 S1, anti-SARS-CoV S, REGN10987, and B38 ( Figure 5). Neither human IgG nor rabbit IgG was significantly detected on the tissues of the pups treated with the

Binding of pathogenic antibodies to fetal and diseased tissues
In order to further evaluate the pathogenicity of the anti-COVID-19 spike antibodies in humans, the REGN10987 with the highest pathogenic potential was tested for antibody binding to various human fetal tissues, or mutiple human diseased tissues from tissue array slides (US Biomax). The results are shown in Figure 6. REGN10987 bound to mutiple human fetal tissues of the lungs, heart, kidneys, brain, pancreas, liver, thymus, and testicles, but not of the esophagus ( Figure 6A). In addition, the REGN10987 also bound to the fetal tissues of retina and coroid, sclera, and eye ball (data not shown). The data indicate that the unmatured fetal tissues are vulnerable to a pathogenic antibody such as REGN10987. In addition, REGN10987 bound broadly to the human inflammatory tissues or cancer tissues of the respiratory, cardiovascular, urinary, and digestive systems ( Figure 6B-6C) Figure 6. Binding of the human monoclonal anti-COVID-19 S1 antibody, REGN10987, to various human fetal tissues (A), and inflammatory and cancer tissues (B-C). SCLC: small cell lung carcinoma, CC: kidney clear cell.
adrenal gland, peripheral nerve, thyroid gland, spleen, adenohypophysis, testicle, prostate, bone marrow, uterine cervix of cancer adjancent normal tissue ( Figure  7). In addition, REGN10987 also bound to the tissues of parathyroid gland, pericardial mesothelium, and adjacent normal sclera of eye (data not shown). The data indicate that certain anti-COVID-19 S1 antibodies, such as REGN10987, are highly pathogenic because it has the high potential to bind to healthy human tissues, activating self-attacking immune responses and inducing serious adverse reactions in vivo. Based on the results, clinical detection of pathogenic antibodies during the COVID-19 infection may be helpful in predicting the consequences of a patient with a serious infection.
Taken together, the in vitro and in vivo data of the current study revealed that certain pathogenic antibodies specific to the COVID-19 spike protein can be the cause of a serious COVID-19 infection, and can cause serious complications from COVID-19 infections. Further, the pathogenic antibodies can bind to unmatured fetal cells or tissues and cause abortions, postpartum labors, still births, and neonatal deaths of pregnant females. It should be noted that the majority (70% or more) of the anti-COVID-19 spike antibodies inducible by the COVID-19 virus are non-pathogenic since the pathogenic antibodies consist of less than 30% of the total number of antibodies, according to the data of monoclonal antibodies in this study.

Discussion
The current study revealed the pathogenic roles and the detailed mechanism of action (MOA) of certain antibodies specific to the spike proteins of coronaviruses such as the COVID-19 virus and the SARS-CoV virus ( Figure 8). We had discovered that in a mouse model, pre-injection of anti-influenza immune sera induced more severe infections than the mice infected with an influenza virus alone 14 . Wang and co-workers reported that anti-SARS-CoV spike antisera promoted SARS infection through antibody-dependent enhancement (ADE) in vitro 22 . Liu and co-workers reported that anti-SARS-CoV spike immune sera induced by a SARS-CoV vaccine caused acute lung injury by promoting MCP1 and IL-8 production and monocyte or macrophage recruitment and accumulation in SARS-CoV infected macaque models 23 . The previously reported mechanism of action (MOA) of these anti-spike antibodies is ADE-based, in that the antibodies enhance viral infectivity 24 . The current study, revealed for the first time, the self-cell targeting MOA of pathogenic antibodies, in which the antibodies bind to host vulnerable cells or tissues and mislead immune responses to attack host-self. Our study also explored a new mechanism of pathogenisis (MOP) of highly pathogenic viral infections. The MOP is caused by the pathogenic antibodies inducible by highly pathogenic viruses such as the COVID-19 virus and SARS-CoV virus. Moreover, the pathogenic antibodies may be related to the cause of infection-related autoimmune diseases, including those in COVID-19 long haulers.

Dualistic roles of anti-viral antibodies
Based on the traditional concept, the antibodies induced by an infectious pathogen are protective to a host because they can neutralize the pathogen and prevent or treat the infectious disease. Nevertheless, the roles of such antibodies can be dualistic. Not wishing to be bound to theory, we noticed that some antibodies can cross react to host cells or tissues and trigger immune reactions to attack the self-cells and self-tissues. The data of the current study showed such pathogenic actions of anti-spike proteins of the COVID-19 and SARS viruses. Sialic acids are predominant components of the mucous membrane at the outer surface of cell membranes and mainly act as biological masks or receptors 25 . Cells or tissues with sialic acid are recognized as "self" 25 . After the loss of sialic acids the cellular structures become "non-self" 25 , which can activate immune responses.
Sialic acids are also an important attachment molecule of receptors for some viruses, such as coronaviruses and influenza viruses 25,21 . During an infection of such viruses, the sialic acid on the infected cells such as lung epithelium cells could be removed or destroyed by the receptor destroying enzyme (RDE) of a coronavirus such as the COVID-19 virus, or the sialidase of influenza viruses. The damaged cells with missing sialic acid on the cell surface become vulnerable to the pathogenic antibodies induced by the virus (Figure 8).
For example, the current study showed that antibodies specific to the COVID-19 spike and the spike glycoprotein of the SARS-CoV virus could significantly bind to the damaged lung epithelium A549 cells ( Figure 1) and kidney embryonic HEK-293 cells with missed sialic acid on the cell surface. The antibody binding could activate and mislead the immune response to attack self and induce the injury of multiple organs in vivo. For example, injection of the antibodies specific to the COVID-19 S1 or the SARS-CoV S proteins to pregnant mice induced fetal and neonatal deaths and the injury of multiple organs of mouse pups born to the dames, as shown in Figures 1-4 and Table 2.
In contrast, the non-pathogenic antibodies of anti-COVID-19 N and anti-SARS N did not induce significant injury in vivo. Neither the anti-COVID-19 N nor the anti-SARS N antibodies significantly bound to the healthy A549 cells nor to the damaged A549 cells ( Figure 1E). The fetal model was selected because many fetal tissues, including the surface glycoproteins, are unmatured and vulnerable to the pathogenic antibodies. Seven naturally occurring human monoclonal antibodies specific to the COVID-19 S1 (S-RBD) protein 15,16,17,18,19,20 were analyzed with the in vitro assay as described above. The results showed that all antibodies except one (85.7%) did not bind to the healthy A549 cells. Neverthless, two (28.6%) antibodies of B38 and REGN10987 significantly bound to the damaged A549 cells ( Figure 1C) and were selected as potential pathogenic antibodies for further confirmation by the in vivo animal model. The results of the virus-free animal experiments showed that the two antibodies alone induced significant fetal and neonatal deaths and the injury of multiple organs of mouse pups born to the dames, as shown in Figures 1-4 and Table 2.
In contrast, another antibody of the seven monoclonal anti-COVID-19 S1 antibodies, CR3022-b6, served as a control antibody and did not induce significant injury in vivo (Figures 1-4 and Table 2). The data indicated that the in vitro assay is useful for rapid screening of potential pathogenic antibodies, and the virus-free animal model is helpful for confirmation of pathogenicity of pathogenic antibodies. It should be noted that the effect of a pathogenic antibody on the mouse fetus could be reduced if the antibody binds to the mother's tissues. A pathogenic antibody can be induced during a highly pathogenic infection, such as the COVID-19 infection.
The discovery of pathogenic antibodies may solve the mystery of the possible MOP of serious infectious diseases, serious complications, and sequela of a viral infection, particularly of a highly pathogenic viral 11 infection such as the COVID-19 infection. This may also explain the cause of cytokine storms and cytokine release syndrome (CRS), and infection-related autoimmune diseases (including those suffered by COVID-19 long haulers 26 ), infection-relating cancers, and other possible disorders inducible by pathogenic antibodies. The diseases or conditions caused by pathogenic antibodies further include abortion, postpartum labor, still birth and neonatal death of pregnant females related to an infection.

A new pathogenic mechanism of viral infections
The in vitro and in vivo data of the current study support a new mechanisms of pathogenesis (MOP) of highly pathogenic viruses such as the COVID-19 virus. The MOP includes: 1) a highly pathogenic respiratory virus such as the COVID-19 virus causes the initial, primary injury such as local inflammation and cellular damage with missed sialic acids of its target organ, such as lungs, typically within week one of the infection; 2) anti-viral antibodies, including pathogenic antibodies, elevate from week one. The pathogenic antibodies bind to the damaged or the inflammatory cells of the target organ (e.g., lungs) ( Figure 8) as well as to other organs with similar injuries (e.g. kidney, brain and heart); 3) the antibody binding activates and misleads the immune response to attack the self cells or tissues, and induces further damage (the secondary injury); 4) the secondary injury persistently adds further damage to the primary injury, creating a snowball effect, and causes serious conditions such as ARDS, cytokine storms, and even death as the antibodies reach peak levels from week one to weeks 2-3 ( Figure 8); and 5) the self-attacking immune responses misled by the pathogenic antibodies can be persistent, and can accumulate after viral clearance and cause autoimmune diseases as long as the antibodies continue to exist. The primary injury is limited, short, and decreases as the virus is cleared, as seen in regular influenza infections. That means the virus itself is not sufficient to cause serious conditions such as ARDS, cytokine storms, and death. On the other hand, the secondary injury caused by the pathogenic antibodies is longer, broader, and additive because antibodies persist much longer than viruses and can bind nonspecifically to other inflammatory tissues or organs besides lungs. Thus, the superposition of the self-attack immune reactions caused by pathogenic antibodies can lead to serious conditions such as ARDS, cytokine storm, and death. The new MOP can explain why the majority of patients with serious respiratory viral infections such as a COVID-19 infection or a highly pathogenic influenza infection died after week one, especially at weeks 2-3 of the infection course 3,27 , as that matches the time period of antibody peak levels. The new MOP can also explain why the majority of severe or lethal infections of the 1918 influenza pandemic happened to the young 27 , because younger people could induce higher levels of anti-viral antibodies, including pathogenic antibodies. Similarly, certain pathogenic antibodies inducible by other infectious pathogens, such as influenza viruses may also cause serious adverse reactions or autoimmune diseases through this MOP (study ongoing).

Cells or tissues vulnerable to pathogenic antibodies
As shown in Figure 6, the highly pathogenic REGN10987 antibody bound to the majority of human fetal tissues and inflamatory tissues, as well as certain cancer tissues. A common feature of those tissues, including the cancer tissues, is that they consist of actively proliferating cells. Therefore, those cells and inflamatory tissues are vulnerable to pathogenic antibodies, and their interaction with the pathogenic antibodies can be the cause of 1) serious infections, particularly highly pathogenic viral infections such as the COVID-19 infection; 2) serious complications of infections such as ARDS, cytokine storm, or cytokine release syndrome (CRS); and 3) infection-relating inflammation such as pneumonia and kidney failure (Figures 1-4). Further, unmatured fetal cells or tissues are vulnerable to pathogenic antibodies ( Figure 6A). Thus, pathogenic antibodies may be the cause of abortions, postpartum labors, still births and neonatal deaths of pregnant females ( Figure 1B and 1C).

Pathogenic antibodies and autoimmune diseases
Many autoimmune diseases are related to viral infections 28 , yet the pathogenic mechanisms have remained unclear so far. The current study discloses, for the first time, that most inflammatory tissues are vulnerable to pathogenic antibodies ( Figure 6B). Chronic inflammation is a feature of most autoimmune diseases. When an infection occurrs, antibodies including pathogenic antibodies are induced and last from months to half a year. The pathogenic antibodies can readily bind to the pre-existing vulnerable inflammatory tissues of an autoimmune disease, and mislead immune responses to attack the body's own tissues. This process can be repetitive and additive as antibodies elevate and induces significant reactions. The occurrence time and pathogenic course of an autoimmune disease caused by pathogenic antibodies can be short or long, depending on the amount of vulnerable cells or tissues and the amount of pathogenic antibodies. The pathogenic antibodies generally persist for months to half a year, and thus the autoimmune disease can occur during that period, especially during the period of antibody peak levels. COVID-19 long haulers have been reported and the causes remain a mystery 26 . The current study provides a possible pathogenic mechanism of COVID-19 long haulers. Despite the virus being cleared in the recovered COVID-19 patients, the anti-viral antibodies remained and could exist for months to half a year or longer. Certain pathogenic antibodies could bind to vulnerable cells or tissues, such as inflammatory cells induced during the COVID-19 infection, and cause persistent adverse reactions such as chronic inflammation of multiple organs. Thus, with the same MOA, pathogenic antibodies can also be responsible for the longer term effects of the COVID-19 infection. For example, a highly pathogenic anti-spike antibody such as REGN10987 can bind to peripheral nerves and may cause abnormalities of vision and taste. Similarly, the pathogenic antibodies inducible by other infectious pathogens such as an avian influenza virus can be responsible for the adverse reactions or autoimmune diseases caused by the infectious. Additionally, in a chronic viral infection (e.g., an HIV infection), pathogenic antibodies can repeatedly stimulate cellular proliferation for a long time and an infection-related cancer can occur if the cellular proliferation loses control.

Better vaccines
It was surprising that when the pathogenic anti-COVID-19 S1 antibodies was mixed with an equal amount of the non-pathogenic anti-COVID-19 N antibodies, the sickness and death rates caused by the antibody mixture was significantly decreased compared to the results of the injected anti-COVID-19 S1 antibody alone (Table 2 and Figure 1C). A similar result was observed with the highly pathogenic REGN10987 as well with an antibody mixture of the REGN10987 and the two non-pathogenic monoclonal antibodies of CR3022-b6 and CC12.3 (Table 2 and Figure 1C). The data suggested that co-existence of non-pathogenic antibodies can reduce the pathogenicity of pathogenic antibodies. Thus, a vaccine capable of inducing multivalent antibodies may be safer, in which at least one kind of antibody is of the non-pathogenic kind that induce less adverse reactions. One example of such a vaccine is the traditional inactivated viral vaccine (e.g. inactivated COVID-19 vaccine) which induces multivalent antibodies specific for multiple antigens of a virus. As another example, a recombinant or mRNA COVID-19 vaccine capable of inducing the antibodies specific to not only the spike protein but also to the nucleocapsid proteins, or a non-spike protein of the SARS-CoV-2 virus may be safer (patent pending).
Despite the mechanism of action being unclear, it is not likely that the non-pathogenic antibodies affected the actions of the pathogenic antibodies through competitive binding since neither the anti-COVID-19 S1 nor the anti-COVID-19 N bind to the same antigens.
Neither the REGN10987 nor the other two non-pathogenic antibodies bind to the same epitope of the S1 protein of the COVID-19 virus. We hypothesize that non-pathogenic antibodies affect the action of pathogenic antibodies through diluting and interrupting the binding of pathogenic antibodies to vulnerable cells or tissues.

Novel clinic interventions based on the new mechanism of pathogenesis
Based on the new MOP of the highly pathogenic viral infection inducible by pathogenic antibodies, novel clinic interventions for treating and preventing a serious condition of the COVID-19 infection may be developed through interrupting the binding of the pathogenic antibodies to host vulnerable cells, tissues, and organs. For example, the following products and approaches may be effective for the treatment of a serious COVID-19 infection: 1) for patients with serious conditions, therapies capable of removing the pathogenic antibodies, such as replacing the patient's plasma with uninfected healthy human plasma, may be effective and should be performed as soon as possible; 2) 13 immunoglobulin products, serum or plasma from not only recovered patients but also from healthy individuals may be helpful for symptom relief through diluting and interrupting the binding of pathogenic antibodies; and 3) pathogen-derived products such as viral antigens or antigen fragments, and synthetic peptides may be effective (neutralizing pathogenic antibodies).
For the diagnosis of a severe COVID-19 infection, determination of the levels of anti-COVID-19 spike antibodies may be important and may helpful to predict the condition of a patient with a serious infection. This is because the virus can become undetectable after one week, but the antibodies induced by the virus can elevate from week one and reach peak levels at weeks 2-3, accompanied by the development of serious conditions such as ARDS, cytokine storms, and death. Higher levels of the anti-COVID-19 spike antibodies may indicate a worsening of the infection.
In summary, our study revealed the roles and a novel MOA of "pathogenic antibodies" in viral infections. In particularly, we explored pathogenic antibodies of highly pathogenic respiratory viral infections, such as COVID-19. The pathogenic antibodies can be induced during an infection, bind to vulnerable cells or tissues such as actively growing cells, and initiate a persistent self-attack immune response, leading to serious conditions including ARDS, cytokine storms, and death. Further, the pathogenic antibodies may also be responsible for infection-related autoimmune diseases, including those experienced by COVID-19 long haulers. Novel clinical interventions for diagnosis and treatment, including improved vaccines based on the new pathogenic mechanisms caused by pathogenic antibodies, can be developed. In addition, the concept and new pathogenic mechanisms explored by this study may also be applicable to other infectious diseases, such as highly pathogenic influenza infections (study ongoing).

Production of human anti-SARC-CoV-2 spike monoclonal antibodies
Seven naturally occurring human monoclonal antibodies specific to the receptor binding domain (RBD) of the spike protein one (S1) of the SARS-CoV-2 virus have been reported by others 15,16,17,18,19,20 . The seven monoclonal antibodies were reproduced for research use only (HuaAn McAb Biotechnology) according to the gene sequences from the NCBI crystal structure site ( Table 1). The features and information of the antibodies are listed in Table 1.
Paired heavy chain and light chain plasmids of anti-Spike human antibody were transiently transfected into HEK-293FE cells at a mass ratio of 1:2 with polyetherimide (Polysciences). Transfected cells were suspension-cultured in a 5% CO2 containing atmosphere at 37°C and the condition medium (Gibco) was refreshed 8 hours after transfection. After 7 days, the supernatants were centrifuged at 10000r per min (rpm) and purified with Protein A prepacked column (Senhui Microsphere Technology). Purified antibodies were quantified by Nanodrop or BCA Quantification Kit (Beyotime). Antibody quality control was performed using an antigen down ELISA. 1 g/ml antigens of COVID-19 Spike-RBD-mFc (HuaAn McAb Biotechnology) were coated on ELISA plate at 4℃ for overnight. The plate was blocked with 1% BSA and coated with anti-spike human antibody as first antibody for 1 hour at 37℃ successively. Uncombined free antibodies were washed with TBST, the secondary anti-human IgG-HRP (Abcam) were added onto the plate, and the plate was incubated at 37℃ for 30 minutes. Binding affinity of the anti-spike human antibodies were analyzed with the signature of OD450 nm absorption.

Anti-COVID19 N protein antibody production
Two months old New Zealand white rabbits were subcutaneously immunized with 250 μg of purified COVID19-N protein ((HuaAn McAb Biotechnology). The rabbits received 3 booster injections at 2-week intervals with 500 g purified COVID19-N protein.
Serum titer of anti-COVID19-N antibodies were validated by an antigen down ELISA after each immunization. Rabbit serum were then harvested, centrifuged at 10,000 rpm for 10 minutes and purified with Protein A prepacked column (Senhui Microsphere Technology). Purified antibodies were quantified by Nanodrop or BCA Quantification Kit (Beyotime).