“Vaccine-Induced Covid-19 Mimicry” Syndrome: Splice reactions within the SARS-CoV-2 Spike open reading frame result in Spike protein variants that may cause thromboembolic events in patients immunized with vector-based vaccines

During the last months many countries have started the immunization of millions of people by using vector-based vaccines. Unfortunately, severe side effects became overt during these vaccination campaigns: cerebral venous sinus thromboses (CVST), absolutely rare under normal life conditions, were found as a severe side effect that occured 4-14 days after rst vaccinations. Besides CVST, Splanchnic Vein Thrombosis (SVT) was also observed. This type of adverse event has not been observed in the clinical studies of AstraZeneca, and therefore led immediately to a halt in vaccinations in several european countries. These events were mostly associated with thrombocytopenia, and thus, similar to the well-known Heparin-induced thrombo cytopenia (HIT). Meanwhile, scientists have proposed a mechanism to explain this vaccine-induced thrombocytopenia. However, they do not provide a satisfactory explanation for the late thromboembolic events. Here, we present data that may explain these severe side effects which have been attributed to adenoviral vaccines. According to our results, transcription of wildtype and codon-optimized Spike open reading frames enables alternative splice events that lead to C-terminal truncated, soluble Spike protein variants. These soluble Spike variants may initiate severe side effects when binding to ACE2-expressing endothelial cells in blood vessels. In analogy to the thromboembolic events caused by Spike protein encoded by the SARS-CoV-2 virus, we termed the underlying disease mechanism the “Vaccine-Induced Covid-19 Mimicry” syndrome (VIC19M syndrome).

Most of these cases displayed thrombocytopenia combined with a rare cerebral venous sinus thromboembolic event that caused the death in one third of affected vaccinated persons in Germany. Many cases of CVST and some cases of splanchnic vein thrombosis (SVT) were reported in the EU drug safety database and in the UK. However, after several days of suspension, the EMA did not see any problem with the vaccine (18th March), and later also the German STIKO suggested administering the AZ vaccine solely to people over 60 years of age, since most cases in Germany affected women between the age of 20 and 60 years. Meanwhile (10th May 2021), the EMA has suggested to also use the Johnson & Johnson vaccine only for people over 60 years.
However, there remained a question regarding the mechanism behind these thromboembolic events.
Thromboembolic events combined with a thrombocytopenia have many similarities with a known phenomenon termed "Heparin-induced thrombocytopenia" (HIT), where these combined effects are induced by the infusion of heparin. The mechanism involves Fcγ-receptor-mediated activation of platelets by autoantibodies against the platelet derived factor PF4, in complex with heparin, and subsequent coagulation events. However, scientists in Greifswald together with several other international scientists were quickly able to demonstrate that the novel vaccine induced thrombosis and thrombocytopenia (VITT) mechanism is not provoked by antibodies directed against the Spike antigen (1,2), which was an important nding, because it rea rmed the safety of all the other vaccines on the market. Although the proposed VITT mechanism explained the thrombocytopenia, the precise trigger behind the thromboembolic events has yet to be resolved. Therefore, we investigated another possible mechanism. As already mentioned above, Covid-19 is caused by SARS-CoV-2, a single-stranded RNA virus of the ß-Corona virus family, encoding 16 non-structural proteins (NSP's 1-16), 8 accessory proteins (ORF3a, 6, 7a, 7b, 8, 9b, 9c and 10) and 4 structural proteins (S, E, M and N) (3). The spike glycoprotein is responsible both for recognition of host cell membrane receptors ACE2 and TMPRSS2, for mediating fusion with the membrane of a host cell (4). Most importantly, the RNA of this virus is translated and replicated only in the cytosol of infected cells, and thus, evolution of coronaviruses has always taken place in the cytosol of permissive cells, and in the absence of processes that are necessarily taking place when nuclear-encoded genes are being transcribed (capping, splicing and poly-adenylation).
Here, we want to draw attention to the process of RNA splicing (5). All genes encoded by our genome have evolved in the presence of intronic sequences which may have played an important role during evolution to safe-guard coding information, but also to shu e genetic information in order to create new cellular functions. With very few exceptions, nuclear encoded genes exhibit intronic sequences and primary transcripts are subject to splice reactions to eliminate intronic sequences. For this purpose, consensus splice donor and acceptor sites have evolved to drive this process that occurs cotranscriptionally in our cell nuclei. The RNA consensus sequences are recognized by RNA/protein complexes (spliceosome) with consists of nearly 100 proteins and speci c U-RNA's (U1, U2, U4, U5 and U6). These RNA/protein complexes bind via speci c RNAs in a complementary fashion to the consensus sequences. The perfect 5'-intronic site is characterized by "G•GUNNGU", while the 3'-end of an intron is characterized by a branch-A nucleotide and the sequence "YYYYYYYCAG•G (the nucleotide in bold and the dots mark the last and rst nucleotide of exonic sequences, respectively). Noteworthy, splice events are in uenced by exonic splice enhancer (6 nucleotides; ESE) and exonic splice silencer sequences (4-18 nucleotides; ESS) which bind splice-activating or -suppressing proteins.
What is the fundamental difference between mRNA and vector-based vaccines? The mRNA vaccines are delivered by a lipid nanoparticle containing the appropriate mRNA molecule -coding for the spike protein of SARS-CoV-2 -to muscle cells surrounding the injection site. Cells that have successfully taken up these nanoparticles will release their cargo mRNA into the cytosol, where it will be translated into Spike protein in the rough endoplasmatic reticulum (ER). Subsequently, the translated and folded Spike proteins will be post-translationally modi ed in the ER and Golgi apparatus and transported to the outer membrane -as membrane-anchored proteins. This way, the immune system is able to recognize the viral antigen, which in turn triggers the initial events for all subsequent immunological processes to produce speci c B-and Teffector cells.
What happens to the same Spike gene when delivered via an adenoviral system? The adenovirus life cycle includes the infection of cells, uncoating of the virus in the cytosol, entry of the adenoviral DNA into the nucleus, and subsequently gene transcription by the host transcription machinery (6). All adenoviral systems follow exactly these steps (Ad5, Ad26 and chimp Ad). Thus, the SARS-CoV-2 Spike gene will be transcribed inside of the nucleus and subsequently exported as mRNA out of the nucleus. Arriving in the cytosol, the mRNA will again be translated into the Spike protein (see above).
And exactly here lies the problem: the viral piece of DNA -deriving from an RNA virus -is not optimized to be transcribed inside of the nucleus. Solely this 3,822 nucleotide long open reading frame, coding for a primary product of 1274 amino acid long Spike protein, contains 6 predicted splice donor and 5 predicted acceptor sites. This problem becomes even more severe when using codon-optimized Spike reading frames (depending on the company: up to 13 splice donor and 11 acceptor sites; see Fig. 1A). Thus, it could well be that the Spike open reading frame of SARS-CoV-2 is potentially disrupted by arbitrary splice events when transcribed inside the nucleus. Most, if not all, of these undesirable splice events would produce shorter protein variants, disrupting the Spike protein upstream of the C-terminally located membrane anchor, and thus, leading to soluble Spike protein variants.
To investigate this hypothesis, we examined a nuclear transcribed Spike open reading frame in a "splice reporter system" in order to qualitatively and quantitatively measure these splice events. In addition, we investigated a Human adenovirus C type 5 (HAdV-C5) encoding for a codon-optimized SARS-CoV2 Spike protein and the AZD1222 vaccine after infecting Hela cells to identify potential splicing events. Here, we present our data and discuss the consequences.

Splice site prediction
The open reading frame of the SARS-CoV2 Spike protein was analyzed by using the SpliceRover online tool (bioit2.irc.ugent.be/rover/ splicerover) which predicts splice donor and acceptor sites by using convolutional neural networks to achieve state-of-the-art accuracy. Here, we excluded all predicted sites below a score of 0.15 to identify only highly signi cant sites (7). In addition, we used the Spike sequences together with the Alternative Splice Site Predictor (ASSP) online tool, just to have another algorithm that identi es potential splice sites (8).

Development of a Spike/splice reporter system
The cloned Spike open reading frame was used to establish an additional splice reporter system. The Spike/splice reporter system consists of the full-length Spike gene including the stop codon, followed by an intronic sequence that contains the necessary branch-A nucleotide, as well as a perfect splice acceptor site that has been fused in three different reading frames (0, +1, +2) to a near full-length Luciferase gene (without start codon). To this end, Luciferase activity can only be measured if splicing occurs. Thus, 3 independent Spike/splice reporter systems were constructed and cloned again into our Sleeping Beauty transposon vector pSBbi-GP (9).

Stable transfection and transduction experiments
All 3 designed splice reporter constructs (pSBbi::Spike-Luc-0/+1/+2-GP; see above) were stably transfected into HEK293T cells and selected for 3-5 days with 2µg/ml Puromycin. Cells were used for Luciferase experiments or for the isolation of total RNA (Qiagen) in order to investigate potential splice events by RT-PCR experiments.
Production and puri cation of HAdv-C5-Spike vectors HAd5::Spike vector particles used in this study are E1and E3-deleted replication-incompetent vector particles based on human adenovirus species C type 5 (based on GenBank AY339865.1, Δ441-3522 and Δ28123-30813). The CMV-promoter controlled SARS-CoV2 Spike expression cassette with SV40 poly-A is based on the GenBank sequence YP_009724390.1, codon-optimized for Homo sapiens and was inserted in the E1 region. Particles were produced in E1-complementing N52.E6 cells (10) and puri ed by CsCl 2 gradient ultracentrifugation as described earlier (11). In brief: 4 x 10 8 cells were transduced with MOI 300 from stock solution. 48 h post transduction cells were harvested, resuspended in 3 ml 150 mM NaCl, 50 mM HEPES, pH 7.4 and lysed by freeze/thaw cycles. Cell debris was separated by centrifugation at 2,000 x g for 10 min, and supernatants were layered on a CsCl step gradient (density bottom: 1.41 g/ml; density top: 1.27 g/ml, 50 mM HEPES, 150 mM NaCl, pH 7.4) and centrifuged at 176,000 x g for 2 h at 4°C. Vector particles were aspirated and further puri ed by a consecutive continuous CsCl gradient (density: 1.34 g/ml, 50 mM HEPES, 150 mM NaCl, pH 7.4) and centrifuged at 176,000 x g for 20 h at 4°C. Vector particles were aspirated and desalted by size exclusion chromatography using PD10 columns (GE Healthcare). Physical vector titers were determined by optical density measurement at OD 260nm (12). To test for splice events in the 3 reporter cell lines, we used the following primer sequences: For analyzing splice events between the Spike and Luciferase ORF, we used the forward primer S.2482.F 5'-CTTGCAGATGC TGGCTTCAT-3', S.2982.F 5'-TAGGTTGATCACAGGCAGACT-3' in combination with the Luc.3912.R 5'-TTC GA GTGGGTAGAATGGCG-3' for all three reporter constructs to monitor the alternative splice events, that create the different Spike-Luciferase fusion proteins.

Page 9/18
The same cell lines were used to measure luciferase activity. Since most splicing events from the Spike open reading frame to the ATG-less Luciferase cassette will destroy the membrane anchor, we also investigated the supernatant of the 3 cell cultures for the presence of Luciferase. Brie y, we seeded 5,000 cells per well in black glass bottom 96-well plates (Greiner BioOne). After 2 days of culture, GFP uorescence (RFU) was measured with the Varioskan Flash Plate-reader (Thermo Fischer Scienti c).
Subsequently, D-Luciferin ( nal concentration of 187.5ug/ml) was added with the automatic dispense system, and Luciferase activity was measured (RLU). A normalization procedure was used to express the RLU/RFU ration to account for intercellular differences. All measurements were carried with n=6, while luciferase activity in medium was measured once at two different time points (t=0 and t=48h). For his purpose 200µl medium was centrifugated for 5 minutes at 800 x g, and 100µl supernatant was used again to measure Luciferase activity.

In silico analyses of potential splice sites in various Spike open reading frames
The Spike open reading frame of the Wuhan SARS-CoV-2 isolate was tested for potential splice sites by using the available SpliceRover online tool. In addition, we used the online Alternative Splice Site Predictor (ASSP) tool to con rm the predicted splice sites (data not shown). Both servers identi ed several dozens of splice sites in the 3,822 nucleotide long open reading frame of the Spike gene, and so we decided to lter these by their individual scores to nd the ones which are highly likely to be used (wildtype Spike gene: 11, see Table 1A). As shown in Figure 1A, we performed the same in silico analysis also for codon-optimized Spike open reading frames. The AZ sequence was deduced from a publication which describes the use of the GeneArt codon optimizer (Fisher Scienti c), while the Johnson & Johnson sequence was obtained directly from the Janssen company (kindly provided by Roland Zahn). Codon optimization resulted in both cases in a higher number of potential splice sites (AZ: 24, Janssen: 20, see Table 1B and 1C).
The SARS-CoV-2 Spike open reading frame enables arbitrary splice events that enables the secretion of a soluble Spike protein First, we analyzed potential splice events within the open reading frame of wildtype Spike, where we had already predicted 6 potential splice donor sites. Such unsaturated splice donor sites may give rise to arti cial splice events (to cryptic splice acceptors) or even generate trans-splicing events. When we tested all three splice reporters in classical Luciferase assay, we realized that the "+2 construct" exhibited the strongest Luciferase activity, while the "0 construct" revealed a much lower activity. The activity measured for the +1 construct was near the values measured non-transfected HEK293T cells which served as negative controls (see Figure 1B). More importantly, we also tried to analyze the medium of the cultured cells and could indeed show that tiny amounts of Luciferase activity could be measured also extracellularly. Due to the design of the 3 splice traps, Luciferase activity can only be measured when splicing occurs. And if splicing occurs within the Spike open reading frame, the C-terminal located domain that serves as membrane anchor will be lost. To this end, these Luciferase experiment strongly indicated that splice events are indeed taking place.
To validate these ndings, we used isolated RNA from all three cell lines and investigated potential splice events by sequencing RT-PCR bands that we obtained when we performed 4 independent RT-PCR experiments (4 Spike primers pointing 3'-prime against 1 Luciferase primer pointing 5'-prime). As exemplarily shown in Figure 1B, we observed several splice reactions from predicted splice sites (e.g. SD1-3), but also internal splice reactions that lead to frame-shift events. We also found 2 splice events downstream of the ACE2 binding domain that fused again to Luciferase. At this point, we have to conclude that splice reactions do indeed take place and all of them cause a C-terminal truncation of the Spike protein, exactly as predicted by our initial hypothesis.
In order to also analyze potential splice events in the vector-based vaccines, an HAd5::Spike constructs and the AZ vaccine were used to transduce Hela cells. After 2 days, RNA was isolated and analyzed again by RT-PCR experiments. This time, however, primers were designed not only for the Spike open reading frame (LK76, LK78, LK113), but also a primer for a gene that is located 3' of the Spike transgene in the backbone of the adenovirus (LK114). As shown in Figure 1C (left panel), some smaller PCR bands can be observed when analyzing the full-length open reading frame of the Spike transgene. Noteworthy, a strong splice donor site is predicted at position +506, and several splice acceptor sites further downstream may explain the observed pattern. The middle panel demonstrates the integrity of the second half of the Spike cassette. In the third panel, a full-length PCR band was observed, as well as 2 smaller bands. These smaller bands might be explained by splice events between the C-terminal portion of Spike with the further downstream localized pIX gene. The difference in size can be explained by a splice donor site at nt 3,614 within the 4,075 shortly upstream of the pIX gene (-14), which was subsequently con rmed by direct sequencing of the faster migrating PCR band. This arti cially spliced transcript can only be explained by transcriptional read through into the pIX ORF, or by a trans-splicing event from the Spike transcript into the 5'NTR of the pIX transcript. Either way, it will lead to a fully functional Spike-pIX fusion protein that could be secreted into body uids due to the missing membrane anchor of Spike.

Discussion
Here, we present rst molecular evidence that vector-based vaccines encoding the Spike protein exhibit a problem that is completely absent in mRNA-based vaccines. This is due to the fact that during the vaccination step, the adenoviral DNA enters the nucleus and use the host machinery to transcribe its (trans)genes inside the nucleus. However, RNA viruses have evolved in the absence of any posttranscriptional modi cation systems that are usually enabled to process the primary RNA transcripts of nuclear encoded genes.
In this study, we experimentally validated our assumption that potential splice events cause the production of Spike protein variants that have lost the important membrane anchor, resulting in secreted, soluble Spike protein variants. We could show that (1) predicted splice donor sites are used to create Spike-Luciferase fusion proteins, that Luciferase activity can be measured intra-and extracellularly which was a clear indication that splice reactions are occurring, and (3) that codon-optimized Spike proteinthat exhibits more and stronger splice donor sites -was able to fuse with the pIX gene either by transcriptional read-through and splicing, or by a trans-splicing event. All these results indicate that splice reactions occur and create soluble Spike protein. Soluble Spike protein has been described to cause adverse effects, e.g. a strong in ammatory response on endothelial cells [13][14][15][16]. Moreover, nearly all severe cases of SARS-CoV-2 infections (COVID-19) suffer from life-threatening thromboembolic events due to the many viruses with Spike surface protein in the blood stream. Even pseudoviruses with Spike protein on the surface cause strong in ammatory reactions in tissues and endothelial cells, indicating the danger of this protein when available in a systemic fashion [17].
Soluble Spike variants together with newly built antibodies against Spike protein as well as the highly speci c blood ow conditions in the central venous sinus of the brain may result in the rare but severe events after vaccination observed with ADZ1222/Vaxzevria. Noteworthy, the vaccine from Johnson & Johnson appears to carry fewer splice donor sequences, especially SD 506 and SD 3614 (see Table 1B and 1C), which are the strongest predicted splice donor sites in the AZD1222 sequence (see Fig. 1A). This may explain the ~ 10-fold lower incidence of severe side effects with the Johnson & Johnson vaccine when compared to the AZD1222 vaccine.
In principle, such thromboses may occur in any site of the human body where endothelial cells express ACE2. Soluble Spike proteins which still exhibit the important core portion of the S1 domain (R319-F551) will be able to bind these receptors. When the immune system starts to produce antibodies against the Spike protein, the endothelial cells will not only bind the soluble Spike protein variants, but would also be decorated with the newly formed antibodies. This will give rise to strong in ammatory reactions either by ADCC (antibody dependent cell-mediated cytotoxicity) or CDC (complement dependent cytotoxicity) occurring in these vessels at various sites where such soluble Spike protein variants accumulate. In ADCC, NK cells could be recruited via their surface receptors (CD16 or CD32) to increase in ammation. During CDC, the C1q protein may recognize the local accumulation of Fc receptors to attract C3b, and thus, will recruit erythrocytes that express the CR1 receptor on their surface or will cause cell damage and in ammation via the normal complement cascade (18).
Several questions are remaining, e.g. why does this happen preferentially in the sinus of the central nervous system? The involvement of central nervous system sinuses (besides other sites in the body) could be explained by the non-unidirectional blood ow (depending on body posture or when sleeping) in these venous vessels that are missing the usual venous valves. Thus, the residence time of soluble Spike protein in this area of our body and the possibility to bind to endothelial cells expressing ACE2 is per se increased. This simply increases the probability of adverse immune reactions at these special places of our venous vessel system. Therefore, we propose a pathological disease mechanism that is depicted in Fig. 1D. On one hand, the recently described VITT mechanism is based on the arti cial activation of PF4 by adenoviral proteins or DNA molecules, which can similarly to heparin, act as a poly-anion to mediate PF4 activation. In patients that exhibit a high load of auto-antibodies against PF4, this may cause the observed thrombocytopenia (Fig. 1D, left side).The other side of the pathological disease mechanism is depicted as well (Fig. 1D, right side). Based on our splicing data, membrane-anchored and soluble Spike protein variants are produced after the vaccination procedure. When the immune system now starts the production of anti-Spike antibodies (days 4-16), these antibodies will recognize the membrane-anchored as well as soluble Spike proteins. However, the soluble fraction of Spike protein variants is disseminated throughout the body and concentrates at various sites of those endothelial cells expressing the ACE2 surface protein. These ACEbound Spike protein variants will become targets of the newly produced antibodies and will cause an ADCC/CDC-mediated in ammatory reaction involving several immune cells. It may also cause the already described NET reaction of neutrophils which has already been shown to induce thromboembolic events (19)(20)(21). It also may cause a strong in ammatory reaction by destroying endothelial cells at those sites, explaining the phenomenon of the rarely observed capillary leak syndrome (Clarkson-disease; 22).
If this mechanism is correct, we may also speculate about some other peculiar observations. What about the age-distribution of thromboembolic events and why this happens more in younger people than in older people? As a matter of fact, elderly people may use more frequently, or even on a daily basis, drugs that decrease blood coagulation and in ammatory events (ASS100 or Vit K inhibitors such as Macurmar). This could be a reason why elderly people suffer much less from these thromboembolic events. In addition, an older immune system displays more immune senescence which means that young people exhibit stronger immune reactions than elderly people, and women even stronger than men. All this would imply a higher incidence in young women when compared to men or elderly people (23,24).
Another speculation could potentially explain why this happens so rarely in rst-vaccinated people. Neutralizing antibodies prevent binding of soluble Spike proteins to endothelial cells that express ACE2 on the cell surface. Moreover, only ACE2-bound soluble Spike protein will cause the above mentioned ADCC reactions. Therefore, it could well be that due to speci c MHC combinations some vaccinated people are unable to produce neutralizing antibodies against Spike. Such an immunological situation in combination with the presence of auto-antibodies against PF4 could potentially explain the rather low incidence of central venous sinus thrombosis -or other thromboembolic events, although all vaccinated persons should face the same pro-in ammatory situation due to the presence of soluble Spike protein when vaccinated with a vector based vaccine.
Based on our ndings, we strongly suggest that the Spike open reading frames -wildtype or codonoptimized -in vector-based vaccines has to be re-optimized to avoid unintended splice reactions and to increase the safety of these pharmaceutical products. Vice versa, all mRNA-based vaccines should represent safe products, because the delivered mRNA will only be translated into surface antigen, without having any possibility to participate in nuclear splice events.

Declarations
Con ict of interests: The authors declare no con ict of interests Figure 1 1A: Splice site prediction within the Spike open reading frames Splice site prediction was carried out by using SpliceRoover. Splice donor sites are given in red, splice acceptor sites in green. SpliceRoover calculates splice sites with a score between 0 to 1, but only splice site with >0,15 were displayed. All sites were displayed in 4 colors as indicated. All splice sites are numbered with "0", "+1" or "+2", to indicate how the open reading frame is disrupted. Therefore, all splice events between splice site with identical