Potential Molecular Mechanisms of Rare Anti-Tumor Immune Response by SARS-CoV-2 in Isolated Cases of Lymphomas

Recently, two cases of complete remission of classical Hodgkin lymphoma (cHL) and follicular lymphoma (FL) after SARS-CoV-2 infection were reported. However, the precise molecular mechanism of this rare event is yet to be understood. Here, we hypothesize a potential anti-tumor immune response of SARS-CoV-2 and based on a computational approach show that: (i) SARS-CoV-2 Spike-RBD may bind to the extracellular domains of CD15, CD27, CD45, and CD152 receptors of cHL or FL and may directly inhibit cell proliferation. (ii) Alternately, upon internalization after binding to these CD molecules, the SARS-CoV-2 membrane (M) protein and ORF3a may bind to gamma-tubulin complex component 3 (GCP3) at its tubulin gamma-1 chain (TUBG1) binding site. (iii) The M protein may also interact with TUBG1, blocking its binding to GCP3. (iv) Both the M and ORF3a proteins may render the GCP2-GCP3 lateral binding where the M protein possibly interacts with GCP2 at its GCP3 binding site and the ORF3a protein to GCP3 at its GCP2 interacting residues. (v) Interactions of the M and ORF3a proteins with these gamma-tubulin ring complex components potentially block the initial process of microtubule nucleation, leading to cell-cycle arrest and apoptosis. (vi) The Spike-RBD may also interact with and block PD-1 signaling similar to pembrolizumab and nivolumab- like monoclonal antibodies and may induce B-cell apoptosis and remission. (vii) Finally, the TRADD interacting “PVQLSY” motif of Epstein-Barr virus LMP-1, that is responsible for NF-kB mediated oncogenesis, potentially interacts with SARS-CoV-2 Mpro, NSP7, NSP10, and spike (S) proteins, and may inhibit the LMP-1 mediated cell proliferation. Taken together, our results suggest a possible therapeutic potential of SARS-CoV-2 in lymphoproliferative disorders.


Introduction
In most cases, cancer is either reported to be a comorbid condition, or associated with COVID-19 disease severity from SARS-CoV-2 infection [1]. However, it was recently reported that a patient with classical Hodgkin lymphoma (cHL) showed disease remission upon infection with SARS-CoV-2 [2]. A similar observation has been reported in a case of follicular lymphoma (FL) [3]. The molecular profiles and the infecting SARS-CoV-2 strains of these patients are unknown, and the authors suggest that the complete remission of the diseases is due to an unknown anti-tumor immune response exerted by SARS-CoV-2 [1,2]. These findings have prompted us to elaborate on the molecular mechanisms behind the SARS-CoV-2 induced remission of cHL and FL.
To place the anti-tumor effect of the SARS-CoV-2 possessing a single-stranded RNA (ssRNA) genome in the right context, it is appropriate to mention that several other ssRNA viruses have demonstrated oncolytic activity, leading to efficient killing of tumor cells while causing no damage to normal tissue [4]. This oncolytic effect is not unique to RNA viruses as it has also been documented for DNA viruses, such as adenoviruses, herpes simplex viruses, and vaccinia viruses [5]. The oncolytic activity of viruses relates to the induction of high expression of protein kinase PKR, the interferon signaling pathway, and the activation of caspases inducing apoptotic cell death [6]. Alphaviruses, measles viruses, rhabdoviruses, and Newcastle disease viruses (NDV) have been used in preclinical animal models and in clinical trials, showing encouraging results for cancer therapy [7]. However, the oncolytic effects of viruses vary. For instance, significantly lower Zinc-finger antiviral protein (ZAP) expression in tumors prevented the viral RNA degradation and translational inhibition of the oncolytic M1 alphavirus commonly occurring in normal tissue [8]. In the case of measles viruses, it has been demonstrated that CD46 mediates virus attachment, entry, and virus-induced cell-to-cell fusion for the MV-Edmonston strain [9]. Typically, the high density of CD46 on tumor cells contributes to extensive cell fusion and enhancement of viral gene expression and provides a compelling explanation for the oncolytic specificity of the MV-Edmonston strain and its favorable use for CD46-targeted cancer therapy. Related to rhabdoviruses, genome-wide RNAi screening identified the endoplasmic reticulum (ER) stress response pathways as important modulators for sensitization to caspase-2-dependent apoptosis and cell death [10]. The tumor-cell-killing mechanisms of the NDVare based directly on the formation of multinucleated syncytia, activation of the extrinsic apoptotic pathway, activation of ER stress pathways, and involvement of MAPK pathways [11]. Indirectly, NDV can induce secretion of proinflammatory cytokines and chemokines, enhanced adhesion of leukocytes, upregulation of MHC, and cell adhesion molecules to activate tumor-specific lymphocytes.
cHL cells exclusively express CD30, and brentuximab vedotin (SGN-35 or Adcetris) is an FDA-approved CD30 targeted drug for the treatment of cHL [18]. SGN-35 is an anti-CD30 monoclonal antibody (mAb) attached to an anti-microtubule compound monomethyl auristatin E (MMAE). Upon binding to CD30, SGN-35 is internalized into cHL cells and releases MMAE that binds to tubulin and thus leads to cell-cycle arrest and apoptosis [18]. However, 20% to 30% of cHL cases were found to relapse [19]. Anti-CD20 mAbs such as rituximab are used to treat FL. Anti-CD19 and anti-CD47 mAbs may also be attractive targets for immunotherapy of FL [20].
Genomic amplification 9p24.1 in cHL has been associated with overexpression of programmed death ligand 1 (PD-L1) and activation of Janus kinase 1 and signal transducer and activator of transcription (JAK/STAT) signaling for disease relapse. Therefore, blocking PD-L1 and PD-L2 receptor PD-1 could be a better treatment option. More recently, T-cell checkpoint inhibitory mAbs such as nivolumab and pembrolizumab against PD-1 have been found to increase the overall survival in cHL [21,22]. Both pembrolizumab and nivolumab bind to the PD-L1 binding site of PD-1 and block the access of PD-L1 and PD-L2, thus preventing the relapse of cHL [21]. FL can also express PD-1, and therefore, PD-1 immune checkpoint blockers such as pembrolizumab can also be promising targeted therapeutics in FL [20].
It is well-established that Epstein-Barr virus (EBV) infection is associated with increased risk, pathogenesis, and immunocompromised conditions in the elderly with cHL [23,24]. EBV-positive FL is not rare, and EBV may play a specific role in disease progression and/or lymphomagenesis [25]. The latent oncogenic membrane protein 1 (LMP-1) of EBV activates NF-κB, JAK/STAT, and PI3K/AKT pathways, leading to apoptosis-prone germinal center (GC) B-cells and prevention of plasma cell differentiation [23]. The cytoplasmic signaling domain of LMP-1 recruits tumor necrosis factor receptor-associated factors (TRAFs) and tumor necrosis factor receptor type 1-associated DEATH domain protein (TRADD) to activate NF-kB signaling-mediated B-cell proliferation [26].
The patient was EBV positive in the cHL case report, although no genetic or molecular profiles were provided [2]. Neither have any information in the case of FL been reported [3]. Therefore, it is difficult to evaluate the underlying molecular mechanism of the disease remission due to SARS-CoV-2 infection. Hence, we considered all possible genetic makeups of these patients. In our first hypothesis (i), we assume that SARS-CoV-2 brings remission in cHL/FL cells using a similar mechanism as brentuximab vedotin (SGN-35 or Adcetris) does. In this process, we presumed that the SARS-CoV-2 might bind to the extracellular domain of any cHL-specific cell surface CD marker using its Spike-RBD to attach and enter into lymphoma cells. Upon internalization, some protein of SARS-CoV-2 may interact with cell division-related proteins to bring apoptosis or cell-cycle arrest or termination of cell division. The second possibility/hypothesis (ii) is that as the patient was positive for EBV, some SARS-CoV-2 protein may interact with the LMP-1 of EBV upon internalization of SARS-CoV-2 into the cell to inhibit the interaction of LMP-1-TRAFs and/or TRADD from abolishing the LMP-1 mediated cell proliferation or cell de-differentiation. In the third hypothesis (iii), we presume that any protein of SARS-CoV-2 may bind to PD-1, similar to the targeted mAb drugs pembrolizumab or nivolumab, and induce remission of cHL/FL.

Protein Structures and 3D Modeling
Crystal structures used in this study are provided in Table S1. Since the SARS-CoV-2 strains of the infected lymphoma patients are unknown, we have used SARS-CoV-2 proteins of the original Wuhan, China strain (GenBank: NC_045512). The crystal structures of human CD30/TNFRSF8 (NP_001234.3), CD15/FUT4 (NP_002024.1), SARS-CoV-2 membrane (M) protein (YP_009724393.1), and LMP-1 (YP_401722.1) of EBV are unavailable. It is likely that the inability to crystallize the CD30, CD15, and LMP-1 of EBV is determined not only by their transmembrane nature but is also associated with high levels of intrinsic disorder predicted in these proteins (Supplementary File S1). Therefore, we used I-TASSER [27] and RaptorX web servers [28] to model these proteins. The models were further refined using GalaxyRefine [29], and the stereochemical quality of the protein structures were determined using the PROCHECK tool available at the SAVES v6.0 server (https://saves.mbi.ucla.edu/ (accessed on10 May 2021)). Finally, models were selected for further analysis based on residues in the most favored region of the Ramachandran plot. The TMHMM Server v.2.0 (http://www.cbs.dtu.dk/services/TMHMM/ (accessed on 10 May 2021)) was used to determine the transmembrane helices and extracellular sequences of all proteins.

Protein-Protein Docking
We performed protein-protein docking using ZDOCK [30] and HDOCK [31] servers. We identified the binding sites/residues from crystal structures using Ligplot+ v.2.2 [32] and from the literature in most cases. However, when we did not have access to such residue knowledge, we performed blind docking. The top ten complexes from ZDOCK or HDOCK were used for further analysis, selecting specific criteria as per the specific objective of the docking (see: Section 3). UCSF Chimera [33] was used to visualize and analyze the crystal structures and the docked complexes.

3D Protein Models
Compared to I-TASSER [27], a better model was obtained using RaptorX [28] for human CD30, human CD15, and the SARS-CoV-2 M protein. The RaptorX derived CD30 model was similar to previously described models [34,35]. Therefore, we used this structure of CD30 for protein-protein docking after refinement using GalaxyRefine [29]. The Ramachandran plot of the refined CD30 structure ( Figure 1A) showed that 87.0% of the residues are in the most favorable regions (Table 1, Figure S1). TMHMM analysis suggested that residues 1 to 385 of the CD30 are exposed to the exterior of the cell and therefore, this area was used in docking experiments. For CD15, the final modeled structure ( Figure 1B) showed a Ramachandran plot with 92.0% residues in the most favorable regions (Table 1, Figure S2). Residues 171 to 530 of CD15 were found on the exterior of the cell as per TMHMM analysis. RaptorX-derived SARS-CoV-2 M protein when refined with GalaxyRefine ( Figure 1C), showed a Ramachandran plot with 92.5% of residues in the most favorable regions (Table 1, Figure S3) and as per TMHMM, the residues 1-19 and 74-77 of the M protein were predicted to be extracellular. While we modeled the LMP-1 of the EBV following the same method, 96% of residues were found in the most favorable regions of the Ramachandran plot ( Figures S4 and S5). However, this refined structure showed five transmembrane helices. According to a previous report [26], the LMP-1 has six transmembrane helices, which we obtained directly from the RaptorX. Therefore, we did not use the GalaxyRefine derived structure of LMP-1 ( Figures S4 and S5) for further analysis and docking but used the RaptorX-derived LMP-1 model, containing six transmembrane helices ( Figure 1D), which showed 87.7% of residues in the most favorable regions of the Ramachandran plot (Table 1, Figure S6). The TMHMM algorithm did not generate any result for LMP-1. Therefore, we considered the membrane topology as described by Kieser, 2007 [26]. All these selected 3D models were used for protein-protein docking. Note: the red color shows the extracellular location and corresponding amino acid residues, green shows the transmembrane location and corresponding amino acid residues, and purple color indicates the intracellular or cytoplasmic location and corresponding amino acid residues.  The first step of SARS-CoV-2 infection is the attachment of the virus to human cells at the human angiotensin-converting enzyme 2 (hACE2) through its S protein receptorbinding domain (RBD), which then facilitates cell membrane fusion for entry into human cells [36,37]. To evaluate our first hypothesis if the S protein binds to any cell surface receptor of cHL/FL, we performed protein-protein docking between the Spike-RBD and reported lymphoma cell surface markers (CD15, CD20, CD27, CD30, CD40, CD45, CD80, CD86, CD95, and CTLA-4/ CD152) using the ZDOCK server [30].
In general, a receptor-binding site of a protein should be in open conformation to bind to its ligand. An in silico structural analysis also predicts that a "1up2down" conformation of the Spike-RBD (one open and two closed monomers of the S protein homotrimer) probably is required prior to binding to the hACE2 [38]. Similarly, the critical residues of the Spike-RBD that interact with the hACE2 are Lys417, Tyr453, Gln474, Ph e486, Gln498, Thr500, and Asn501 [39]. However, any peptide that inhibits the interaction of the Spike-RBD and the hACE2 must bind to three key positions of the Spike-RBD (Gly485/Phe486/Asn487, Gln493, and Gln498/Thr500/Asn501) [40]. To verify and standardize the ZDOCK-or HDOCK-based Spike-RBD and human CD receptor docking, we used the crystal structure of the Spike-RBD-hACE2 complex (PDB: 6LZG, the structure of novel coronavirus spike receptor-binding domain complexed with its receptor hACE2, Method: X-Ray diffraction, Resolution: 2.50 Å) and characterized the interacting residues using Ligplot+ v.2.2. The results showed that the Spike-RBD interacts with the hACE2 through Lys417, Tyr449, Ala475, Asn487, Gly496, Gln498, Thr500, and Gly502 residues forming 10 H-bonds ( Figure S7). While we combined these results and our additional structure-based analysis, we observed that four regions of the Spike-RBD-(i) R1-Lys417, Tyr449, Tyr453, and its adjacent residues; (ii) R2-Gln498-Tyr502; (iii) R3-Gln474, Ala475, Phe486, Asn487; (iv) R4-Gly496, Gln493-residues could be critical for strong interaction with any receptor. R2 and R3 are probably the most essential regions ( Figure S8).

The SARS-CoV-2 M and ORF3a Proteins Interact with the Human Gamma-Tubulin Complex Components and May Inhibit Tubulin Nucleation
During microtubule (MT) polymerization, the gamma-tubulin ring complex is initially formed as a gamma-tubulin small complex. In this process, each of the GCP2 and GCP3 molecules recruits one TUBG1 molecule at its top and then forms a heterodimer binding through their lateral positions to start the MT nucleation process. Alpha-and beta-tubulin form heterodimers, and through alpha-tubulin, these heterodimers bind to gamma-tubulin. The other GCPs (GCP4, 5, and 6) also follow the same process, and finally, the microtubule polymerization starts [41,42].
As part of our first hypothesis, next, we focused on identifying whether any SARS-CoV-2 protein interacts with the tubulin complex-forming components and blocks the microtubule formation to arrest the cell cycle. According to Gordon et al. [43], the SARS-CoV-2 M protein interacts with the human gamma-tubulin complex component 2 (TUBGCP2/ GCP2) and gamma-tubulin complex component 3 (TUBGCP3/GCP3). Similarly, Chen et al., reported that SARS-CoV-2 ORF3a potentially interacts with GCP2, GCP3, and even gamma-tubulin (TUBG1) [44]. However, in none of the cases, the authors have provided any further details about the binding sites and the effect of these interactions.
Therefore, if the M or ORF3 protein inhibits MT nucleation to restrict the cell cycle, it should bind to any of these three possible areas: (i) the sites where the GCP2 and GCP3 bind laterally to form heterodimers (ii) the gamma-tubulin binding sites of GCP2 or CGP3, and (iii) the TUBG1 sites that are used to interact with GCP2 or GCP3.
To understand this interaction, first we identified the binding residues between GCP2 and GCP3. For this purpose, we used two crystal structures: the human GCP2-GCP3 complex (PDB: 6V6B) and the human gamma-tubulin ring complex (PDB: 6V6S). Next, we standardized the residue positions of PDB: 6V6S corresponding to PDB: 6V6B, and we noted the interacting residues in the GCP2-GCP3 complex according to PDB: 6V6B using Ligplot+ v.2.2. We identified the interacting residues between GCP2-GCP3 at lateral positions as GCP2: Ser283, Gly396, Glu242, Gln350; GCP3: Arg252, Ser323, His343, Gln529, which form three H-bonds and one salt bridge ( Figure 3A and Figure S28). We found that the chains C/G (GCP2) and B/F (GCP3) of PDB: 6V6S show similar interactions as the interactions found in PDB: 6V6B.  Figure S36). Table 3. Interacting human microtubule ring complex protein residues (bold) with SARS-CoV-2 protein residues (Roman).

The M and ORF3a Proteins May Interact at GCP2-GCP3 Lateral Binding Sites
To explore if the M and ORF3a proteins hinder the lateral binding of GCP2 and GCP3, we selected the GCP2 (GCP2: Ser283, Gly396, Glu242, Gln350) and GCP3 (GCP3: Arg252, Ser323, His343, Gln529) interacting residues and docked them with the M and ORF3a proteins using the HDOCK server. Among the four residues of GCP2 that form three Hbonds with GCP3, the M protein may interact with two residues (Ser283 and Gln350) along with two nearby residues, forming six H-bonds (complex-7) (Tables 3 and S3A, Figures 3B and S33). In complex 3, GCP2 interacts with the M protein, forming nine H-bonds (Thr235, Ser243, Arg290, Ser393) (Tables 3 and S3A, Figure S34). However, the residues are not the same as the ones that interact with GCP3. When we performed the docking between the SARS-CoV-2 M protein with GCP3 targeting its GCP2 binding residues, only one residue (Arg252) of GCP3 interacted in three different complexes (complex-4, -6, -7). However, other interacting residues of GCP3 are not also close to the mapped interacting residues involved in GCP2-GCP3 interaction (Table S3B).
We performed a similar analysis with the ORF3a protein as we did for the M protein. Although the ORF3a protein interacts with GCP2 with a maximum of seven H-bonds (complex 4), it does not involve any residue of GCP2 that interacts with GCP3 (Tables 3 and S4A, Figure S35). On the other hand, the ORF3a protein potentially interacts with GCP3 at its GCP2 binding site. In complex 7 it makes five H-bonds, and one salt bridge involves two residues (out of four) of GCP3 (Ser323, Arg252, His343-salt bridge) that are involved in lateral binding with GCP2 and the other two residues (Gln322, Arg333) (Tables 3 and S4B, Figures 3C and S36). In complex 3 it makes seven H-bonds and involves seven residues of GCP3; however, only one residue (Arg252) interacts with GCP2 (Tables 3 and S4B, Figure S37).
Taken together, the M protein may not bind to GCP3 at its GCP2 interacting site. However, the M protein possibly interacts with GCP2 at its GCP3 binding site using two residues (out of four) ( Table 3, Figures 3A,B, S28 and S36). On the other hand, the ORF3a may interact with GCP3, blocking at least three (out of four) of its GCP2 binding residues (Tables 3 and S4B, Figures 3A,C and S36). Therefore, the potential interactions of M-GCP2 or ORF3a-GCP3 may affect the GCP2-GCP3 lateral binding and hinder microtubule nucleation.

Both the M and ORF3a Proteins May Interact with GCP3 at Its TUBG1 Binding Sites
To understand if the M and ORF3a proteins interact with TUBG1, we docked both proteins at four different positions: (i) TUBG1 binding sites/residues of GCP2, (ii) TUBG1 binding sites/residues of GCP3, (iii) GCP2 binding sites/residues of TUBG1, and (iv) GCP3 binding sites/residues of TUBG1.
Out of 10 GCP2 residues that interact with TUBG1, the M protein can maximally bind to three such GCP2 residues (Asp554, Arg711, Asp561) and two additional residues (Tyr723, Ser885) in complex 5, forming six H-bonds (Table 3 and Table S3C, Figure S38). In complex 1, the interactions show five H-bonds. However, only 3 GCP2 residues (Glu731, Cys684, Ser885) interact with TUBG1 (Tables 3 and S3C, Figure S39). The M protein interacts with GCP3 using maximally 10 H-bonds involving three key residues (Lys671, Lys689, Ser709) and six other residues of GCP3 in complex 1 (Tables 3 and S3D, Figures 4A-C and S40). In complex 5, the M protein binds to three TUBG1 interacting residues of GCP3 (Asp572, Arg681, Ser709) and four other residues (Gly571, Phe612, His716, Asn883), which are close to the TUBG1 interacting residues of GCP3 forming eight H-bonds (Tables 3 and S3D, Figure S41). The ORF3a protein binds to the TUBG1 interacting GCP2 residues in two complexes (complex 3 and 9), forming seven H-bonds. In complex 9, seven residues are involved where four residues (Gln719, Cys684, Asn716, Asp561) interact with TUBG1 (Tables 3 and S4C, Figure S42). However, in complex 3, only four residues of GCP2 interact with the ORF3a protein, where three residues (Cys684, Gln719, Asn890) of GCP2 interact with TUBG1 (Tables 3 and S4C, Figure S43). In GCP3 (TUBG1 binding site) -ORF3a interaction, in complex 1, GCP3 interacts with the ORF3a protein involving seven residues and 12 H-bonds, where three residues of GCP3 (Arg681, Ser709, Gln717) interact with TUBG1 (Tables 3 and S4D, Figures 4D-F and S44). In complex 3 we found that five Hbonds and only two such GCP3 residues (Asn609, Gln717) interact with the ORF3a protein (Tables 3 and S4D, Figure S45). Therefore, our results suggest that the SARS-CoV-2 M protein may interact with the TUBG1 binding site of GCP3 and probably block GCP3 binding to TUBG1. However, considering the number of H-bonds and the number of GCP residues involved, the ORF3a protein may bind to GCP3 at its TUBG1 binding site more strongly than GCP2. The probable 3D interactions are shown in Table 3 and Figure 4A-F. Since both the M and ORF3a proteins are probably interacting with GCP3 at its TUBG1 binding sites, we also performed a superimposition analysis that also shows the M and ORF3a protein interaction with GCP3 may inhibit TUBG1 binding to GCP3 (Table 3, Figure 4G,H). Nevertheless, additional analyses are needed if these interactions of the M and ORF3a proteins inhibit TUBG1 binding to GCP3.

The M but Not the ORF3a Protein May Strongly Interact with TUBG1 at Its GCP3 Binding Sites
In docking studies of the M protein with TUBG1 binding sites that interact with GCP2, no interaction involving the GCP2 binding residue of TUBG1 was detected except in complex 1, where only one matching residue (Tyr248) was observed. However, in this model, we found nine H-bonds (Tables 3 and S3E, Figure S46). Similarly, in complexes 7 and 8, the M protein interacted with TUBG1 using 9 and 10 H-bonds, respectively (Table 3 and Table S3E, Figures S47 and S48). For TUBG1 binding sites for GCP3, we got better results. The M protein can bind with nine H-bonds (complex-1) and involves five residues (Tyr248, Asp252, His334, Arg341, Trp351) of TUBG1 (out of 14) that interact with GCP3 (Tables 3 and S3F, Figures 5A-D and S49). In complexes 8 and 9, the M protein interacts with TUBG1 making 10 and 7 H-bonds, respectively. In complex 8 it uses two common residues (Pro162, Arg265), and in complex 9 four common residues (His334, Arg341, Ser355, Gln357) (Tables 3 and S3F, Figures S50 and S51).  Table 3. (A) Superimposed view of the M protein (spring green) binding to TUBG1 (purple) at its GCP3 binding position, blocking GCP3-TUBG1 interaction (corresponding to complex 1, Figure S49 In the ORF3a protein, although it makes seven H-bonds in complex 3, it does not involve any TUBG1 residue that interacts with GCP2. However, in complex 1, only four H-bonds are formed by four residues, where only two residues (Tyr248, Gln357) of TUBG1 are involved in binding with GCP2 (Tables 3 and S4E, Figures S52 and S53). In the case of GCP3, we also found the same trend. Out of 16 TUBG1 residues, that interact with GCP3, only four of them (Tyr248, His334, Arg341, Gln357) probably bind to the ORF3a protein, forming four H-bonds in complex 1 (Tables 3 and S4F, Figures 5E-H and S54). In this complex, His334 and Arg341 form H-bonds with the ORF3a protein. However, in TUBG1-GCP3 interaction, these two residues of TUBG1 form salt bridges. In complex 3, a total of six residues of TUBG1 (Arg212, Asp216, Val305, Arg343, Arg390, Gln394) bind with the ORF3a protein making seven H-bonds. However, none of these TUBG1 residues interact with GCP3 (Tables 3 and S4F, Figure S55).
Taken together, our analysis suggests that the M protein may bind to TUBG1, blocking its binding to GCP3 more strongly than GCP2. The probable 3D interactions between TUBG1-M are shown in Tables 3 and S3F, Figures 5A-D and S49. Considering the number and specific residues of TUBG1 involved in GCP2 and GCP3 interactions, it is unlikely that the same residues interact with the ORF3a protein and interfere with TUBG1 binding to GCP2 and GCP3 (Tables 3 and S4F, Figures 5E-H and S54).

LMP-1 Signaling Domains of EBV Potentially Interact with SARS-CoV-2 Proteins
The LMP-1 protein of EBV is essential for EBV-associated oncogenesis. Since the cHL patient described by Challenor and Tucker [2] tested positive for EBV, we assess in our second hypothesis that some SARS-CoV-2 proteins might inhibit the oncogenic signaling of LMP-1 by binding to its cytoplasmic signaling motif residues 204-208 aa/"PQQAT" and 379-384 aa/"PVQLSY" that recruit TRAFs and TRADD, respectively, to activate the oncogenic NF-kB signaling-induced B-cell proliferation [26].
To test this second hypothesis, we performed a SARS-CoV-2 proteome-wide docking with our modeled LMP-1 using ZDOCK. In this docking, we selected the LMP-1 residues 202-210 aa (TRAFs binding site) and 375-386 aa (TRADD interacting site) and docked them with SARS-CoV-2 proteins without selecting any residues. We considered few principles in this analysis, (i) since the LMP-1 interacting motifs of TRAFs or TRADD are only 4-5 amino acids of length, a SARS-CoV-2 protein must interact with at least two critical amino acids of any of these motifs, and (ii) since more than two ZDOCK models should show the same interacting amino acids of TRAFs or TRADD binding motifs with hydrogen bonds.

Does the Spike Protein Interact with PD-1 to Block Access to PD-L1 and PD-L2?
Finally, we checked our third possible mechanism, i.e., whether the SARS-CoV-2 Spike protein interacts with PD-1 and acts similarly to the T-cell checkpoint inhibiting monoclonal antibodies (mAbs) like nivolumab and pembrolizumab to induce PD-1 mediated apoptosis.
Although the Spike-RBD does not bind exactly to the pembrolizumab and nivolumab interacting residues of PD-1, it may interact with PD-1 at very close proximity to the PD-1 residues that interact with these two mAbs. Therefore, the Spike-RBD may bind to PD-1 and may act as pembrolizumab and nivolumab-like mAbs to block access to PD-L1 and PD-L2 leading to cHL cell apoptosis and remission.

Discussion
During the initial days of the COVID-19 pandemic, there were no reports of any direct correlation between SARS-CoV-2 and lymphoma. However, in December 2020, based on a multi-omics approach, we predicted that lymphoma would have an association with COVID-19 [47,48]. In January 2021, Challenor and Tucker first brought to our attention that a 61-year-old man with cHL and who tested positive for EBV showed complete remission of cHL after he was infected with SARS-CoV-2 [2]. The second case was reported by Sollini et al. in February 2021, where a 61-year-old patient suffering from FL showed complete remission upon SARS-CoV-2 infection [3]. The molecular mechanisms behind the SARS-CoV-2 induced remission of these lymphoma cases are yet to be understood.
Challenor and Tucker suggested that SARS-CoV-2 may induce an unknown specific anti-tumor immune response mechanism [2] that may also be responsible for a previously reported case of spontaneous regression of diffuse large B-cell lymphoma (DLBCL) after infection with Clostridium difficile [49]. Similarly, Sollini et al. [3] suggested that the SARS-CoV-2-induced remission of their reported FL case was probably due to a "flare phenomenon" as observed in immunotherapy that finally results in an "abscopal effect".
A recent report in May 2021 from Spain also suggested that SARS-CoV-2 triggers an anti-tumor immune response in lymphoma [50]. In addition, a 22-year-old Hodgkin lymphoma patient suffering from COVID-19 was successfully treated with the PD-1 inhibitor pembrolizumab [51], and the cHL was generally treated with CD30 targeting brentuximab vedotin [18]. Therefore, like the anti-tumor mechanisms of other oncolytic ssRNA viruses [4,[8][9][10][11], we also focused on possible anti-tumor immune responses by SARS-CoV-2 in the reported cHL and FL cases, hypothesizing three possible mechanisms. As per our results, the proposed overall mechanisms of SARS-CoV-2-induced remission of cHL and FL cases are presented in Figure 7.
ACE2 is expressed in lymphomas [52] and SARS-CoV-2 infection is associated with increased severity and mortality in lymphomas [53]. However, these SARS-CoV-2-induced remissions of cHL and FL are very isolated events [2,3] compared to global cHL and FL cases and the magnitude of the COVID-19 pandemic. Additionally, the molecular profiles of these patients [2,3] are unknown. Therefore, it is not likely to be a general mechanism to explain this remission phenomenon. Our first hypothesis-based results show that, the initial contact and attachment of SARS-CoV-2 to cHL or FL cells in these two cases may be possible through the interactions between the SARS-CoV-2 Spike-RBD and cell-surface markers for cHL or FL such as CD15, CD27, CD45, and CD152 (Table 2, Figure 2A-D).
Previous reports suggest that some unclassified FL expresses CD15 [54] and CD27 [55]. Expression of CD152 is also reported in cHL [12,56]. CD15 is also expressed in some specific groups of cHL patients [57,58] and its expression can predict the disease outcome [59]. Circulating blood cells in cHL patients show clonal expression of CD27 [60]. CD15 is also required for cell adhesion to platelets to promote cell proliferation and migration beyond the lymphatic system [61]. Therefore, CD15 may be a potential target in cHL. CD27 activates protein kinase C (PKC) and induces cellular proliferation in B-cell lymphomas [62] and anti-CD27 mAb shows antitumor activity [63]. CD45 is also aberrantly expressed in certain cases of lymphomas [64,65], and phosphatase activity of CD45 is required for lymphoid cell proliferation [66]. Inhibition of CD45 phosphatase activity negatively regulates Src family tyrosine kinase (SFK) signaling and thereby induces G2/M cell-cycle arrest and cell apoptosis [66]. Therefore, CD45 could be a possible therapeutic target in lymphoma. Similarly, activation of CD152/ CTLA4 induces cell proliferation and tumor growth in lymphomas through activation of the TYK2-STAT3 pathway [67], and anti-CD152 mAb ipilimumab could be a potential targeted therapeutic for lymphomas [67,68]. PD-1 is upregulated in lymphomas, a hallmark of EBV-associated lymphoproliferative disorders, and this overexpression is associated with disease relapse [69,70]. Dual targeting of PD-1 and CD152 using combinations of anti-PD-1 and anti-CD152 mAbs may be a potential therapy in relapsed lymphoid malignancies [71]. Based on these facts, if we assume that these two patients [2,3] express any of these very specific receptors such as CD15, CD27, CD45, and CD152 that are not commonly expressed in cHL and FL, and that the SARS-CoV-2 Spike-RBD may bind to any of these receptors like a mAb and thereafter works in a similar fashion to brentuximab vedotin (SGN-35 or Adcetris), this suggests a possible molecular mechanism of the remission of the lymphoma patients as described by Challenor and Tucker [2] and Sollini et al. [3]. As the expression of CD15, CD27, CD45, and CD152 is also associated with lymphoma cell proliferation these CD molecules can be therapeutic targets [61][62][63][66][67][68]. If the Spike-RBD interaction inhibits the CD expression, it may also induce cell-cycle arrest and apoptosis.
Our analysis found that the Spike-RBD can potentially interact with CD15, CD27, CD45, and CD152 expressed by lymphoma cells (Table 2, Figure 2A-D). As per our first hypothesis, after attachment to lymphoma cells through CD15, CD27, CD45 or CD152, the SARS-CoV-2 enters into lymphoma cells by an unknown mechanism and upon cellular entry, it will use some of its proteins to initiate cell-cycle arrest or apoptosis by interacting with cell-cycle machinery components. Gordon et al. [43] and Chen et al. [44] reported that the M and ORF3a proteins of SARS-CoV-2 interact with the gamma-tubulin complex components GCP2, GCP3, and TUBG1. The binding of GCP2-TUBG1, GCP3-TUBG1, and GCP2-GCP3 is essential for the microtubule nucleation process [41,42]. Our analysis suggests that the M protein may bind to GCP2 at its GCP3 binding site ( Figure 3B), and the ORF3a protein may bind to GCP3 at its GCP2 binding residues (Table 3, Figure 3C); thus, both the M and ORF3a proteins can potentially render the GCP2 -GCP3 lateral binding. Further, the M and ORF3a proteins may also bind to GCP3 at its TUBG1 binding site (Table 3, Figure 4A-H). Additionally, we also predicted that the M and ORF3a proteins might bind to TUBG1, blocking its interaction with GCP3 (Table 3, Figure 5A-H). Therefore, if the M and ORF3a proteins are involved in such interactions, the microtubule nucleation process will stop leading to cell-cycle arrest.
Some coronavirus infections induce apoptosis in host cells that may be required for viral replication and propagation in respective hosts [72]. SARS-CoV and SARS-CoV-2 induce apoptosis through distinct mechanisms [73]. SARS-CoV uses the 7a protein to induce apoptosis, activating the p38 MAPK pathway [74]. SARS-CoV also uses the S, M, and N proteins for induction of apoptosis [75,76]. Recently, the SARS-CoV-2 ORF3a protein was reported to induce apoptosis via the release of cytochrome C [77]. As per our prediction, the ORF3a protein may also be involved in cell-cycle arrest at a very early stage, leading to apoptosis or cell death ( Table 3, Figures 4D-F and 5E-H).
PD-1 immune-checkpoint-blocking mAbs such as pembrolizumab and nivolumab are used in cHL [21,22] and FL [20] treatment to induce apoptosis. Although we do not know if the reported cHL and FL cases [2,3] are PD-1 positive, we analyzed whether the SARS-CoV-2 Spike protein may act like pembrolizumab and nivolumab, which bind to PD-1 to block the access of PDL-1 and PDL-2 and thus downregulate PD-1 signaling and induce apoptosis. Our analysis observed that the Spike-RBD potentially interacts with PD-1 at the binding sites of these two mAbs, although only a few overlapping residues of PD-1 that interact with pembrolizumab/nivolumab are found to interact with the Spike-RBD (Table 4, Figure 6E,F). Therefore, we suggest further validating whether the SARS-CoV-2 Spike protein can also induce apoptosis through negative regulation of PD-1 mediated signaling as per the results of our second hypothesis.
Our previous report predicted that the SARS-CoV-2 infection pathway is involved in crosstalk with other viral pathways, including EBV [47]. Co-infection of EBV in COVID-19 patients is not uncommon, and in COVID-19 patients, reactivation of EBV has been reported, which may be associated with disease severity and other symptoms of long COVID-19 [78][79][80][81]. Additionally, patients with lymphoproliferative disorders showing immunodeficiency and post-traNSPlantation patients subjected to immunosuppression, the synergistic action of EBV and SARS-CoV-2 may increase the fatality rate [82]. The cHL patient of Challenor and Tucker was EBV positive [2], and the LMP-1 of EBV is the main oncogenic protein of EBV that activates oncogenic signaling through activation of NF-κB, JAK/STAT, and PI3K/AKT pathways through its cytoplasmic TRAFs and TRADD binding motifs [23,26]. Contrary to the reported potential synergistic association between EBV and SARS-CoV-2 [78][79][80][81][82], in our analysis (as per our third hypothesis), we found that the 3CL pro /M pro , NSP7, NSP10, and S (Table 4, Figure 6A-D) proteins of SARS-CoV-2 may interact at the TRADD binding sites of LMP-1 thus blocking the access of TRADDs to LMP-1, and this interaction may inhibit the LMP-1-mediated NF-kB oncogenic signaling to induce remission. However, further investigations are required to validate these interactions and their outcomes in lymphoma remission.

Conclusions
This study has explored the possible molecular mechanisms behind the rare phenomenon of lymphoma remission upon SARS-CoV-2 infection. We have focused on a specific anti-tumor immune response caused by SARS-CoV-2. We are currently planning to conduct in vitro experimental validations of our results and if our proposed mechanism is proven right, SARS-CoV-2 may be engineered for effective therapeutic interventions against certain lymphomas and other proliferative disorders.