key: cord-0996400-ji0h2rcq
authors: Peter, Antonia Sophia; Roth, Edith; Schulz, Sebastian R.; Fraedrich, Kirsten; Steinmetz, Tobit; Damm, Dominik; Hauke, Manuela; Richel, Elie; Mueller-Schmucker, Sandra; Habenicht, Katharina; Eberlein, Valentina; Issmail, Leila; Uhlig, Nadja; Dolles, Simon; Grüner, Eva; Peterhoff, David; Ciesek, Sandra; Hoffmann, Markus; Pöhlmann, Stefan; McKay, Paul F.; Shattock, Robin J.; Wölfel, Roman; Wagner, Ralf; Eichler, Jutta; Schuh, Wolfgang; Neipel, Frank; Ensser, Armin; Mielenz, Dirk; Tenbusch, Matthias; Winkler, Thomas H.; Grunwald, Thomas; Überla, Klaus; Jäck, Hans-Martin
title: A pair of non-competing neutralizing human monoclonal antibodies protecting from disease in a SARS-CoV-2 infection model
date: 2021-04-16
journal: bioRxiv
DOI: 10.1101/2021.04.16.440101
sha: a28064747fb5b79027e5d4b6aba33c69ed7cc1ae
doc_id: 996400
cord_uid: ji0h2rcq

TRIANNI mice carry an entire set of human immunoglobulin V region gene segments and are a powerful tool to rapidly generate human monoclonal antibodies. After immunizing these mice against the spike protein of SARS-CoV-2, we identified 29 hybridoma antibodies that reacted with the SARS-CoV-2 spike protein. Nine antibodies neutralized SARS-CoV-2 infection at IC50 values in the subnanomolar range. ELISA-binding studies and DNA sequence analyses revealed one cluster of clonally related neutralizing antibodies that target the receptor-binding domain and compete with the cellular receptor hACE2. A second cluster of neutralizing antibodies binds to the N-terminal domain of the spike protein without competing with the binding of hACE2 or cluster 1 antibodies. SARS-CoV-2 mutants selected for resistance to an antibody from one cluster are still neutralized by an antibody from the other cluster. Antibodies from both clusters markedly reduced viral spread in mice transgenic for human ACE2 and protected the animals from SARS-CoV-2 induced weight loss. Thus, we report two clusters of potent non-competing SARS-CoV-2 neutralizing antibodies providing potential candidates for therapy and prophylaxis of COVID-19. The study further supports the use of transgenic animals with human immunoglobulin gene repertoires in pandemic preparedness initiatives.

Since December 2019 [1] , SARS-CoV-2 has rapidly spread throughout the world, leading to more than 133,971,287 confirmed cases of COVID-19 and 2.902.493 million deaths [2] . Several vaccines have been developed and licensed in worldwide efforts at an unprecedented speed [3, 4] . In countries or age groups with high vaccine coverage, case counts have decreased substantially [5, 6] . For all licensed vaccines, the vaccine antigen is the ectodomain of the spike protein. Its receptor-binding domain (RBD) interacting with the cellular receptor human Angiotensin-converting enzyme 2 (hACE2) as well as its N-terminal domain (NTD) were identified as primary targets of the neutralizing activity of convalescent sera and monoclonal antibodies [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] . Clinical development of some of these monoclonal antibodies is ongoing, already providing evidence of their utility in therapeutic trials [20] [21] [22] . Some have since been licensed for the treatment of COVID-19 (reviewed in [23] ).

After crossing species barriers, zoonotic viruses adapt to new species. Early after the new entry, adaptation to efficient transmission is most likely the most substantial selective pressure, explaining the rapid emergence of SARS-CoV-2 variants, such as those carrying the D614G mutation [24] or more complex mutational signatures in the spike protein [25, 26] . The higher the percentage of a population is that has overcome the first infection, the more robust the selective advantage of variants escaping the immune responses raised against the initial founder virus will be. Evidence that immune escape may already contribute to the emergence of new variants has been reported for the South African B.1.351 and the Brazilian P.1 and P.2 variants [27] [28] [29] . Since the antibody response to current mRNA vaccines precisely mimics the antibody response to natural infection [30] , the vaccines' efficacy against these variants may be impaired [27, [30] [31] [32] . Similarly, the neutralizing potency of monoclonal antibodies may be reduced or completely lost [15, 33, 34] .

Concerning the loss of vaccine efficacy against emerging variants, it has been argued that mRNA and viral vector vaccines can be rapidly adapted to include the spike proteins of the variants of concern. While this can be expected to work for individuals with no prior exposure to SARS-CoV-2 or the COVID-19 vaccines, the outcome is less predictable in the presence of pre-existing antibody responses to the founder spike protein. As variants only differ in a few epitopes, immunizations with spike proteins of emerging variants in individuals previously exposed to the founder spike protein may lead to preferential boosting of pre-existing non-neutralizing antibodies to epitopes conserved between the two spike proteins. More targeted vaccine approaches employing structure-based designs may be needed but require reliable animal models for validation and optimization. Transgenic mice in which the coding sequences for variable regions of the immunoglobulin (Ig) murine heavy (H) and light (L) chains were replaced by coding sequences of variable regions of human H and L chains could be such a model.

Here we report the rapid isolation of fully human SARS-CoV-2 neutralizing monoclonal antibodies from immunized TRIANNI mice harboring the complete human antibody repertoire (patent US 2013/0219535 A1; Fig 1A) . The neutralizing antibodies target the RBD or the NTD of the spike protein, providing a pair of non-competing human monoclonal antibodies for clinical development. Further characterization of the antibodies´ human VH gene usage and divergence from germline ancestor sequences supports the concept of mouse models with a human antibody repertoire in pandemic preparedness efforts and suggests the use of these models in structure-based optimization of vaccine antigens against emerging SARS-CoV-2 variants.

Based on the published nucleotide sequence of SARS-CoV-2 [35] , the first immunogens available early during the current pandemic were expression plasmids encoding the spike protein, the primary target of neutralizing antibodies to coronaviruses [36, 37] . As DNA vaccines also induce good T cell responses, we selected DNA vaccines as our priming immunogens. In addition to a DNA vaccine encoding the wild type S protein (SARS-CoV-2-S DNA), we generated a DNA vaccine in which the coding region of the cytoplasmic domain of SARS-CoV-2 was replaced by the cytoplasmic domain of the G protein of VSV (SARS-CoV-2-SG DNA). For SARS-CoV, it was shown that such a modification increases S protein expression at the cell surface and induces higher neutralizing antibody responses after immunization with exosomal vaccines [38] . The time interval between the DNA priming immunization and the booster immunization allowed for the production and purification of a recombinant S protein stabilized in the pre-fusion state (SARS-CoV-2-S protein) and an exosomal vaccine presenting a membraneanchored form of the SARS-CoV-2-S protein (SARS-CoV-2-S Exo) (Fig. 1B) .

TRIANNI mice were selected for these immunization experiments (Fig. 1A ) because they produce antibodies with a human Ig variable (V) region and murine constant (C) regions. One unique feature of the TRIANNI mice is that the intronic regions between the exons of the variable segments of Ig genes are essentially of murine origin. This design aims to maintain proper regulation of expression and splicing of the V regions of the immunoglobulin exons as a prerequisite for proper affinity maturation. After selection of antibodies of interest, the human V region gene segments can be recombinantly fused to human C region gene segments allowing expression of fully human antibodies for clinical applications.

On February 14, 2020, groups of two TRIANNI mice were either immunized by intramuscular electroporation with the SARS-CoV-2-S DNA and or the SARS-CoV-2-SG-DNA (Fig. 1B) . Three weeks later, the SARS-CoV-2-S DNA primed mice were either boosted with the SARS-CoV-2-S DNA or the SARS-CoV-2-S protein adjuvanted with MPLA liposomes. The SARS-CoV-2-SG DNA primed mice were boosted with SARS-CoV-2-SG or SARS-CoV-2 exosomes at weeks 3 and 4, respectively (Fig. 1B) . Using a flow cytometric assay with cells transiently transfected with SARS-CoV-2-S DNA [38] , antibody responses were analyzed two weeks after the priming and the boosting immunization (Fig. 1C ). Antibodies to SARS-CoV-2-S were detectable in all four animals after the DNA priming immunization and further increased after the booster immunizations.

Sera of the immunized mice were also analyzed for competition with the binding of the cellular receptor hACE2 to the SARS-CoV-2-S protein (Fig. 1D, E) . Two weeks after the booster immunization, all four sera reduced hACE2 binding. Priming with SARS-CoV-2-S DNA seemed to result in slightly higher competition activity. Sera from the mice primed with SARS-CoV-2-S DNA and boosted with SARS-CoV-2-S protein or SARS-CoV-2-S DNA also neutralized wild type SARS-CoV-2 at a 1:100 dilution by 50% and 27%, respectively ( Fig 1F) . Competitive inhibition of hACE2-Fc binding to trimeric S protein by sera (1:200) from control mice and mice at the indicated time points after the first immunization. The mean percentage of binding as compared to control binding is shown (two experiments each performed in triplicates). (F) For the neutralization assay, Vero-E6 cells were infected with the SARS-CoV-2 isolate MUC-IMB-1 in the presence or absence of week 5 sera. SARS-CoV-2 infection was quantitated after 20 to 24 hours by staining with purified IgG from a convalescent COVID-19 patient and a fluorescence-labeled anti-human IgG using an ELISPOT reader. The mean and SEM of triplicates of one experiment are shown.

The TRIANNI mouse (M2) with the best neutralizing activity was boosted with the adjuvanted SARS-CoV-2-S protein on week 6 and sacrificed five days later to generate hybridoma lines. A 1:100 dilution of the serum from this time point neutralized wild-type SARS-CoV-2 by 86%. Cells from the spleen and draining lymph nodes were fused with Sp2/0-Ag14 cells, and hybridoma cells were grown in 106 96well plates seeded at a density of 1,000 cells per well. Between ten to 20 days after the fusion, antibodies secreted by the hybridoma cells were screened by flow cytometry for binding to the S protein. A total of 29 hybridomas could be identified that secreted antibodies binding to the S protein.

These antibodies were designated TRES antibodies (TRIANNI-Erlangen anti SARS-CoV-2-Spike). The screening results of ten representative hybridoma supernatants containing TRES antibodies are displayed in Fig. 2A (for a list of all binding antibodies and their characteristics, see Table 1 ). Intracellular staining revealed that most hybridomas expressed the IgG2c subtype.

The binding antibodies were further characterized in an hACE2 competition assay. Three (TRES6, 224, 567) of the 29 binding antibodies were able to outcompete hACE2 binding very efficiently (Fig. 2B) , indicating that these antibodies might bind the RBD and possibly neutralize SARS-CoV-2. Subsequently, all supernatants were tested in a virus neutralization assay with SARS-CoV-2 MUC-IMB-1, an early isolate from the Webasto cluster in Bavaria [39, 40] . Supernatants from the hybridomas competing with ACE2-binding (TRES antibodies 224, 567 and 6) strongly neutralized SARS-CoV-2 ( Fig.  2C ) and indeed, an RBD-specific ELISA confirmed the RBD-binding of these three antibodies (Fig. 3A) . Three additional hybridoma supernatants could be identified to inhibit virus infection by at least 40% (TRES49, 328, 1293), although they do not compete with hACE2 for binding to the SARS-CoV-2-S protein. They also do not bind to the RBD, recombinant S1 proteins or the S2 protein ( Fig. 3A-C) . TRES49, 328, 1293 bind to a stabilized S protein trimer and recombinant NTD (Fig. 3D, E) . A similar binding profile was also observed for the TRES antibodies 618, 219 and 1209. Thus, these six antibodies differ in their binding profile from TRES6, TRES224 and TRES567. We also identified one hybridoma supernatant (TRES1) that strongly binds to S2 (Fig. 3B) . 

Based on the screening results, hybridomas were subcloned, and antibodies from the supernatants of hybridoma subclones of the TRES antibodies 6, 224, 567, 49, 219, 328, 618, 1209, 1293 and 1 were purified and further characterized. Since the TRES antibodies 6, 224, and 567 compete with hACE2 for binding to RBD ( Fig. 2A) , the 50% effective concentration (EC50) was determined in an hACE2 competition ELISA t0 range from 0.75 to 1 nM ( Fig. 4 A) . Surface plasmon resonance (SPR) analyses further revealed KD Values of purified RBD and NTD binding antibodies against the S protein stabilized in the prefusion state from 9.5 nM to 1.6 nM (Table 1 ). The neutralizing activity of purified antibodies was determined against a second wild type SARS-CoV-2 isolate, CoV-2-ER1 (Fig. 4B ). In contrast to the MUC-IMB-1 isolate, CoV-2-ER1 contains the D614G mutation dominating the SARS-CoV-2 populations in Europe and the US [41, 42] . Antibodies that compete with hACE2 (TRES6, 224, and 567) had IC50s ranging from 64-102 ng/ml, while the IC50s of TRES49, 328, 618 and 1293 were lower, being between 5 and 34 ng/ml. Although the initial screening of hybridoma supernatants had not detected neutralization by TRES618, 219, 1209, the antibodies had an IC50 of 18 ng/ml, 5 ng/ml and 11 ng/ml, respectively. The S2-binding antibody TRES1 did not display neutralizing activity (Fig. 4B ). When the antibodies were assessed in a cell-cell fusion assay, in which Vero-E6 and HEK-293T cells expressing the S protein were allowed to fuse, the RBD-binding antibodies seemed to inhibit cell-cell fusion more efficiently than the neutralizing antibodies not binding to RBD (Fig. S1 ). In general, cell-cell fusion inhibition required higher antibody concentrations and total inhibition of cell-cell fusion was not achieved (Fig. S1 ).

Sequencing of productive immunoglobulin variable regions expressed by hybridomas encoding TRES antibodies revealed a broad usage of V regions of H and L chains with a preference for L chains (Table  1 ). Compared to the corresponding germline sequences, TRES VH and VL nucleotide sequences showed similarities as low as 94.44% and 97.85%, respectively, indicating the presence of somatic mutations and suggesting affinity maturation. At the same time, clonally related hybridomas could be identified. Consistent with their similar binding profiles (Fig. 3, 4) , TRES6, 224, and 567 H and L chain exons contain the same VHDJH and VLJL recombination joining sequences and are therefore designated cluster 1 antibodies. All three antibodies express the VH3-33*01 or VH3-33*06 and KV1D-17*01 gene segments (Table 1 ). An alignment of amino acid sequences of VH and VL of cluster 1 antibodies with the inferred germ line encoded antibody revealed an identical L chain with 5 non-silent somatic mutations in the antigen bindings loop (complementary determining regions, CDR) and one in the framework (FR) (Fig. S2 ) TRES6 and 567 utilize the same VDJ H chain exon with four somatic mutations in the framework. TRES224 contains two additional mutations in the CDR1 and CDR2 and an unusual change of a tyrosine that is conserved in 98% of all human VH segments. A second cluster comprises six TRES antibodies (Table 1 ; TRES49, 219, 328, 618, 1209 and 1293) that share the VHDJH and VLJL recombination joining sequences. Various somatic mutations are also present in the CDRs as well as in the FR of their VH and VL segments (Fig S2) . The non-neutralizing antibodies of a third cluster bind to S1, but not to RBD, and are composed of an H chain with a human VH and a L chain with an unusual mouse VλX region (Table 1 ) [43] , which could be explained by the fact that the mouse VλX region is still present in TRIANNI mice. 

Gene segments encoding the human V exons of the H and L chains of all neutralizing TRES antibodies were fused with the human 1 H and the  L chain constant regions, respectively. Recombinant antibodies entirely composed of human Ig regions were produced in HEK-293F cells and purified by affinity chromatography. As expected, all recombinant TRES antibodies bound to the S protein in the flow cytometric assay (Fig. 5A ).

To explore a potential overlap of epitopes between TRES antibodies of cluster 1 and cluster 2, cells expressing the SARS-CoV-2-S protein were incubated with TRES224hu or TRES618hu in the presence of increasing concentrations of TRES224, 567, 49, 328, 618 and 1293 hybridoma antibodies containing mouse C regions (Fig. 5B, C) . No competition between the cluster 1 and cluster 2 antibodies could be observed. In contrast and as expected, binding of the recombinant TRES224 with human C regions (TRES224hu) was efficiently blocked by TRES567 and 224 (Fig. 5B) , indicating that these two neutralizing antibodies from cluster 1 recognize the same epitope. Accordingly, the binding of TRES618hu was blocked by TRES49, 328, 618 and 1293, verifying that cluster 2 antibodies bind the same epitope. Most importantly, however, the humanized TRES antibodies also neutralized SARS-CoV-2 with similar IC50s as the parental hybridoma TRES antibodies (Fig. 5D) , confirming that the identified antibody sequences confer neutralization. 

Neutralization assays (Fig. 6A ) and spike protein binding assays (Fig. S3) were also performed for two emerging SARS-CoV-2 variants of concern and three antibodies from each cluster. The IC50s of the cluster 1 antibodies against the B.1.1.7 variant were in the range of 6 to 49 ng/ml (Fig. 6A , left panel) and were similar to those for the wild-type virus (Fig. 5B ). IC50s consistently below 10 ng/ml (Fig. 6A , right panel) indicate that the B.1.351 isolate might be more susceptible to neutralization by cluster 1 antibodies than the wild-type virus (see Fig. 5D ). In contrast, cluster 2 antibodies lost their neutralizing activity against both variants of concern (Fig. 6A) .

To further explore the possibility for de novo emergence of resistant variants to cluster 1 and cluster 2 antibodies, the CoV-2 ER-1 isolate carrying the D614G mutation was passaged in increasing concentrations of TRES6 and TRES328. After 5 passages, neutralization assays were performed. While the passaged virus became resistant to antibody neutralization when the antibody was present during passaging, the virus remained sensitive to the TRES antibody when the antibodies were not included during passaging (Fig. 6B) . Whole-genome sequencing of the passaged SARS-CoV-2 viruses revealed the emergence of a TRES6 antibody escape variant with an I68R mutation in the NTD and a T478K mutation in the RBD (Fig. 6C, displayed in pink) . In contrast, the TRES328 escape variant harbored a L241-Y248 deletion with a phenylalanine insertion in the NTD (Fig. 6C, displayed in blue) . Additional point mutations and deletions were observed in proximity to the S1-S2 cleavage sites. These mutations probably reflect adaptation to the replication in the Vero-E6 host cells [44] since these sites were also deleted in the virus variant (P5) passaged in the absence of a neutralizing antibody (Fig. 6C , displayed in white). Additional mutations were detected in the P5 variants at the positions H245 and P812 probably also reflecting viral adaptation to cell culture.

To assess whether the antibodies generated in this study can confer protection from disease, we evaluated the efficacy of one antibody from each cluster in a stringent hACE2 transgenic mouse [45] viral challenge model under post-exposure prophylactic settings. Three groups of mice (n=12/group) were infected intranasally with 300 FFUs of SARS-CoV-2 (MUC-IMB-1). One day later, mice were injected intravenously with 5.25 mg/kg body weight of the cluster 1 antibody TRES6, the cluster 2 antibody TRES328, or the isotype-matched control antibody TRES480. Six animals from each group were sacrificed on day 4 or 10 or according to humane endpoints, and viral loads were determined in the lung, BAL, brain, liver and spleen. Both antibodies reduced the amount of viral RNA in the lung and BAL samples approximately 30-to 200-fold after four days (Fig.  7A ) or approximately 100-o 450-fold after 10 days (Fig. 7B ). Viral load in the other organs was reduced close to the level of detection. Interestingly, the titer of infectious virus in the BAL samples from TRES6 and TRES325 treated animals was below the detection level and, therefore, at least 1000-fold lower than in mice receiving the control antibody (Fig. 7C) . Compared to the isotype control, TRES480, treatment of mice with TRES6 and TRES328 prevented the loss of body weight induced by the viral infection (Fig. 7) and reduced clinical symptoms assessed by a clinical score (Fig. 7E) . Importantly, none of the TRES antibody-treated mice reached clinical endpoints requiring euthanasia, while 4 of 6 SARS-CoV-2-infected mice receiving the isotype control antibody TRES480 had to be euthanized (Fig. 7F ). 

By immunizing TRIANNI mice with a DNA prime protein boost regimen, 29 monoclonal antibodies were obtained that bind to the S protein of SARS-CoV-2 expressed on the cell surface of transfected HEK-293T cells. The use of immunization and screening approaches based on the expression of the wildtype S protein avoids the need for prior knowledge of domains targeted by neutralizing antibodies and allows to obtain a panel of antibodies binding to different epitopes. Indeed, different binding patterns were observed, including antibodies that i) bind to the RBD and compete with ACE2 binding, ii) bind to the S1 subunit without binding to RBD, iii) only bind to the ectodomain of the S protein stabilized in a trimeric form and the NTD and iv) bind to the S2 subunit. In those mice that we first challenged with SARS-CoV-2 and subsequently treated with the monoclonal antibodies, we could also assess any potential disease-enhancing properties. This issue has been suggested as a potential outcome of infections occurring in the presence of vaccine induced antiviral antibodies not providing sterilizing immunity [46] . However, no evidence for such an enhancing effect could be observed. A combination therapy with an antibody from cluster 1 and cluster 2 seems especially attractive since this may extend the neutralization breadth and enhance the genetic barrier for the emergence of antibody-resistant SARS-CoV-2 variants. This assumption is further strengthened by the observation that the B.1.1.7 and the B.1.351 variants maintain neutralization sensitivity to at least one of the two clusters of antibodies. In addition, virus escape mutants from one antibody of each cluster are still sensitive to a representative TRES antibody from the other cluster.

Most of the neutralizing antibodies in convalescent COVID-19 patients are close to the germline sequence [47] . In contrast, the TRES antibodies had mutations in the VH and VL regions including the CDRs (Fig. S. 2) suggesting that affinity maturation can indeed occur in TRIANNI mice after a DNA prime protein boost immunization regimen. Although mice encoding a full [11] or partial [48] human immunoglobulin repertoire have been used previously to generate monoclonal antibodies targeting the spike protein of SARS-CoV-2, direct evidence for affinity maturation, in the context of SARS-CoV-2, has not yet been provided. Using a three-week immunization regimen consisting of a single intradermal priming immunization with DNA encoding the S protein and booster immunizations with an RBD-Fc fusion protein in VelocImmune TM mice, Hansen et al. showed induction of RBD-specific antibody responses and also reported the isolation of neutralizing RBD-binding antibodies from B cells that were sorted based on RBD binding [11] . Wang et al., performed sequential immunizations of H2L2 TM (Harbour Antibodies) mice with spike ectodomains of different human coronaviruses and derived a monoclonal antibody that cross-neutralized wildtype SARS-CoV-2 moderately with an IC50 of 570 ng/ml [48] .

Although extensive affinity maturation does not seem to be necessary for the formation of potent neutralizing antibodies against SARS-CoV-2 [47] , this may be relevant for elicitation of neutralizing antibodies against other pathogens as exemplified by HIV-1 [49] [50] [51] . As observed for four of seven other potent neutralizing antibodies targeting the NTD and recovered from convalescent patients [7] , our TRIANNI mouse-derived NTD antibodies also use the VH1-24 gene segment. Therefore, transgenic animal models supporting affinity maturation of antibodies with human variable regions should be an essential part of future pandemic preparedness efforts and be explored in structure-guided vaccine antigen design against SARS-CoV-2.

Besides the success in generating potent neutralizing antibodies that also confer protection in the hACE2 mouse model, a key criterion for applying TRIANNI mice in future pandemic responses is the time needed to obtain such antibodies. The immunization schedule we used took seven weeks until hybridomas were generated. Growing the hybridomas, their screening and subcloning took approximately six weeks. Sequence analyses from these hybridomas and gene synthesis for the generation of fully human antibodies took another six weeks. Accelerating the development time would certainly be desirable. Whether shortening the immunization schedule still results in potent neutralizing antibodies is difficult to predict and may depend on the degree of affinity maturation needed to generate such antibodies. However, instead of generating hybridomas, antigen-specific memory B cells or germinal center-selected surface IgG-positive plasmablast could be stained with fluorescence-labeled antigen and sorted by flow cytometry. Paired VH and VL sequences from single cells could then be amplified by PCR and directly cloned into expression plasmids, essentially as described for memory B cells from convalescent COVID-19 patients [10, 52] . Transiently expressed recombinant antibodies could then be screened for binding and neutralization and sequenced, probably shortening the development time by several weeks. Immunization of TRIANNI mice and the establishment of SARS-CoV-2-specific hybridoma clones. Two 9-12-week old TRIANNI mice were primed with pCG1-CoV-2019-S ( [53] designated SARS-CoV2-S DNA) encoding wild type SARS-COV-2 spike protein (position 21580 -25400 from Genbank NC_045512). Two additional TRIANNI mice were primed by SARS-CoV-2-SG DNA encoding a chimeric protein in which the intracytoplasmic domain of SARS-CoV-2-S was replaced by the intracytoplasmic domain of VSV-G as previously described for SARS-CoV [38] . The DNA vaccines were delivered by intramuscular electroporation as previously described [54] . Briefly, a total of 30 µg DNA diluted in 60 µl PBS were injected in both hind legs under constant isoflurane (CP-Pharma, Burgdorf, Germany) anesthesia. Immediately after the intramuscular injection, electrical impulses were applied to the injection site with a TriGrid electrode array (provided by Drew Hannaman, Ichor Medical Systems, Inc., San Diego, USA). Mice were boosted intramuscularly either with i) 5 µg of the S protein of SARS-CoV-2 stabilized in a pre-fusion conformation (designated SARS-CoV-2-S protein) and adjuvanted with 25 µg Monophosphoryl Lipid A (MPLA) liposomes (Polymun Scientific GmbH, Klosterneuburg, Austria) into the hind leg, ii) by electroporation of the DNA vaccines used for priming, or iii) with exosomes purified from HEK-293T cells transiently transfected with SARS-CoV-2 DNA as described previously [38] . Details for the immunization schedule are given in Figure 1B . Mice were bled by punctuation of the retroorbital sinus using non-heparinized capillaries (Hirschmann Laborgeräte, Eberstadt, Germany). Serum was collected, inactivated for 30 min at 60°C and kept at -20°C for long-term storage.

One of the immunized mice received a second booster immunization with SARS-CoV-2-S protein at week 5. Five days later, spleen and inguinal lymph nodes were harvested and fused with the azaguanine-resistant SP2/0 Ag14 hybridoma line (ATCC#CRL-1581) using the PEG (Sigma Aldrich, Taufkirchen, Germany) method [55] . Briefly, 2x10 8 spleen cells and inguinal lymph node cells (about 1x10 8 B cells) were mixed with 10 8 Sp2/0 cells and washed 3 times with RPMI1640 medium (Gibco, Thermo Fisher Scientific, Waltham, USA) without FCS. 2 ml PEG was added dropwise within one minute to the suspended cell pellet. The mix of fused cells was resuspended in R10+ medium (RPMI 1640 with 10% FCS, 0.05 mM mercaptoethanol, 2 mM L-glutamine, 1 mM sodium pyruvate, and 1% penicillin/streptomycin; (all from Gibco, Thermo Fisher Scientific, Waltham, USA); enriched with HAT (Gibco, Thermo Fisher Scientific, Waltham, USA) and OPI Media supplement (Sigma Aldrich, Taufkirchen, Germany). Cells were seeded into 96-well plates with approximately 1x10 4 cells per well. 10-20 days post-fusion, hybridoma clones were tested for spike-binding antibodies. Positive clones were expanded in R10+ supplemented with HT (Sigma Aldrich, Taufkirchen, Germany) and subcloned by the limiting dilution method.

To detect antibodies that bind to SARS-CoV-2-S protein, HEK-293T cells were co-transfected with SARS-CoV-2-S DNA or pCG1-SARS-Fra1-S (CoV-1; position 21492 to 25259 Genbank AY291315SARS-CoV) and a GFP reporter plasmid (e.g., pEGFP-C1) using the PEI method as described previously [56] . Briefly, For ELISA-based hACE2 competition, 96-well microtiter plates were coated overnight at 4°C with 20ng/well (400 ng/ml) recombinant RBD that was purified from the culture medium of transfected HEK-293F cells. hACE2 and RBD proteins were generated by transient transfection of HEK-293F cells.

For this the following plasmids were transfected: The plasmid pCEP4-hACE2ecto-Fc encodes a fusion protein containing the N-terminal ectodomain of hACE2 (AA 18-738 NCBI RefSeq: NP_001358344.1) the Hinge-CH2 -CH3 region of human IgG1 and a C-terminal Myc-tag and 6xHis-tag. The vector encoding pSARS-CoV-2-RBD was constructed by inserting a commercially synthesized RBD fragment (AA 319-541, GenBank: QHD43416.1) with an N-terminal IgKappa signal sequence (METDTLLLWVLLLWVPGSTG) and C-terminal TwinStrep-and 3xFLAG-tags (GeneArt, ThermoFisher, Regensburg, Germany) into a pcDNA3.4(+) vector backbone. SARS-CoV-2 RBD and hACE2 proteins were purified from HEK-293F supernatants via affinity chromatography with Strep-Tactin®XT 4Flow agarose (IBA, Göttingen, Germany) following the manufacturer's instructions. Eluted proteins were dialyzed against PBS in Slide A_Lyzer Dialysis Cassettes (Thermo Fisher Scientific, Waltham, USA) and stored in aliquots in 50% glycerol at -20 o C. Protein purity was assessed by SDS PAGE followed by Western blot or Coomassie staining as described previously [57] . Protein concentrations were determined by OD at 260 nm and verified by a Bradford assay (Thermo Fisher Scientific, Waltham, USA). Plates were washed with PBS/0.05% Tween-20 and blocked with 275 µl per well of PBS with 2% BSA at room temperature (RT) for 1-3 hours. Competitional binding in the ELISA set up was achieved by applying 50 µl/well of 0.25 µg/ml human ACE2-Biotin (Acro Biosystems, Beijing, China, # A011-214) followed by 50 µl/well serially 2-fold pre-diluted TRES antibodies (2 µg/ml start concentration). This was incubated at RT for 1-2 hours. Wells were washed, 50 µl/well HRP-coupled Streptavidin (0.25 µg/ml, Merck Millipore, Darmstadt, #OR03L) were added and incubated at RT for 1-2 hours. After another washing step, HRP-bound Streptavidin was detected by adding 50 µl/well TMB substrate (BD Bioscience, Heidelberg, Germany, #555214). The reaction was stopped using 50 µl/well 0.5 M H2SO4 solution, and the OD at 450 nm was determined in a Spectra Max 190 (BMG Labtech, Ortenberg, Germany). EC50 values were calculated by plotting hACE2 activity against antibody concentrations and applying a 4-parameter curve fit using GraphPad Prism 7.02 (San Diego, USA).

Detection of SARS-CoV-2 spike binding antibodies by ELISA. 96-well microtiter plates were coated over night at 4°C with 100ng per well of recombinant NTD (Sino Biological; Beijing, China #40591-V49H), SARS-CoV-2 Spike S2 or S1 Glycoprotein (Virion\Serion, Würzburg, Germany), RBD or a S protein stabilized in a prefusion state (both affinity purified from the supernatant of HEK-293F cells, as described previously [58, 59] ). The plates were washed and blocked with 5% skimmed milk for 1 hour at room temperature and then incubated with purified TRES antibodies or hybridoma supernatants for 1 hour. Next, the plates were washed, and goat-anti-mouse HRP (Dianova, Hamburg, Germany; #115-035-146) was added and incubated for 1hour. The plates were washed, and RLU (relative light units) were detected with the multilabel plate reader Victor X4 (Perkin Elmer, Hamburg, Germany).

Antibodies were captured on a Protein G Chip (Cytiva Lifesciences Protein G Chip; Marlborough, USA) to reach a response level of 500 RU. Following a kinetic titration was performed (tAssociation = 180 s, tDissociation = 360 s) using a three-fold serial dilution of the S protein and spanning 5 concentrations. The chip surface was regenerated using 10 mM glycine, pH 1.5, as recommended by the manufacturer. The data was analyzed with Langmuir Kinetics and the BIAcore x100 Evaluation Software 2.1. The measurements were performed in triplicates on a Cytiva Lifesciences BIAcore X100 Plus. The S protein used was stabilized in a prefusion state as described [58] . It contained an N-terminal TPA signal peptide and a trimerization foldon followed by a His Tag at the C-terminus. The S protein was purified from transiently transfected HEK-293F cells as described previously [59] .

Cloning and sequencing of Ig V exon sequences. According to the manufacturer instructions, RNA was isolated from hybridoma cell lines with the RNAeasy Mini Kit (Qiagen, Aarhus, Denmark). 500 ng total RNA was used with the Template Switching RT Enzyme Mix (NEB Ipswich, USA) to generate 5' RACE cDNA. Briefly, 1 µM of dT40 VN reverse primer and 1mM dNTP Mix were incubated for 5 minutes at 70°C together with 500 ng RNA in a 6µl reaction. The RT Enzyme mix and buffer were added to the reaction together with a Template Switch Oligo containing 3' riboguanosines (rGrGrG; for primer sequences see Supplemental Table 1 ). The reverse transcription/template-switching reaction was incubated for 90 min at 42°C and inactivated at 85°C for 5 min. The template cDNA was then PCR amplified with 5' primers annealing to the template switch oligo and 3' primers specific for the respective Ig constant region of  and  and for VH/V/Vwith Q5 Hot Start polymerase (NEB, Ipswich, USA). PCR products were run on agarose gels, and 500-700 bp bands were excised and purified (QIAquick Gel Extraction Kit, Qiagen, Aarhus, Denmark). The purified PCR products were then sequenced with gene-specific primers or cloned with the CloneJET PCR Cloning Kit (Thermo Fisher Scientific, Waltham, USA) and introduced into competent DH5α E. coli cells. For each PCR product, at least four clones were sequenced to determine a consensus sequence. The consensus sequences were analyzed using VDJsolver [60] and IMGT/V-Quest [61] . [62] variants of concern were obtained from a patient in München or Frankfurt, respectively. Following they were propagated as described previously for the CoV-ER1 variant.

The infectious titers were determined by limiting dilution in 96-well plates. Three days after infection, culture supernatants were removed from the wells and cells were washed with PBS and fixed with 4% paraformaldehyde in PBS for 20 min. Afterward, they were permeabilized for 15 min with 0.5% TritonX in PBS and blocked with 5% skimmed milk diluted in PBS for 1 hour. Subsequently, they were stained with protein G purified sera from a convalescent patient diluted 1:100 in PBS containing 2% skimmed milk. After 1 hour the cells were washed, and a goat anti-human IgG FITC (Jackson ImmunoResearch, West Grove, USA #109-096-088) antibody was applied. After 1 hour and a washing step, positive wells were identified by a CTL-ELISPOT reader (Immunospot; CTL Europe GmbH, Bonn, Germany). The signal was analyzed with the ImmunoSpot® fluoro-X™ suite (Cellular Technology Limited, Cleveland, USA) software, and TCID50s were calculated as described previously [63] .

Virus neutralization assay. Neutralizing activities of hybridoma supernatants, antibodies and sera were assessed in a microneutralization assay based on detecting the number of virus-producing cells via immunofluorescence. For this, 1x10 4 Vero-E6 cells were seeded per well of a 96-flat bottom plate one day before the infection. Purified antibodies, sera or hybridoma supernatants were first diluted and pre-incubated for one hour with 1,88x10E 4 infectious units of SARS-CoV-2 stocks per well in a final volume of 100 µl. All dilutions of hybridoma supernatants or purified antibodies were prepared in cell OptiPRO TM culture medium. The dilution provided specifies the end-concentration of the antibodies after mixing with the diluted virus stock. After removing the cell culture medium, seeded Vero-E6 cells were incubated for one hour with 100 µl per well of the pre-incubated virus-antibody mix. Afterwards, the supernatant was discarded, the cells washed once with 100 µl PBS and 100 µl fresh cell culture medium added. The cells were incubated for 20-24 hours, washed with PBS, fixed with 4% paraformaldehyde, and subsequently stained as described above. Plates treated with murine antibodies were stained with protein G purified sera from a convalescent patient and plates treated with human antibodies with an antibody mix directed against SARS-CoV-2 at a concentration of 250ng/ml. IC50 values were calculated by plotting the virus activity in percent against the antibody concentrations and using the normalized response vs. inhibitor equation (variable slope) or inhibitor vs. response-Variable slope (four parameters) of GraphPad Prism 7.02.

Quantitative antibody competition assay. For the assessment of the quantitative antibody competition SARS-CoV-2-S DNA transfected HEK-293T cells were incubated with 100 µl of a humanized protein G purified TRES antibody at a concentration of 250 ng/ml and serial dilutions of TRES antibodies with a murine Fc region at final concentrations ranging from 2.5 µg/ml-0.01 ng/ml. The cells were incubated for 30min on ice, washed, and bound antibodies were detected with a mouse IgG2a Alexa647-conjugated antibody directed against human IgG-Fc (BioLegend, San Diego, USA #409320). After washing the mean fluorescence intensities of transfected cells were determined with an Attune Nxt (Thermo Fisher Scientific, Waltham, USA) and the Flow Logic software TM (IInivai Technologies, Mentone, Victoria, Australia).

Generation of escape mutants. 5x10 6 Vero-E6 cells were seeded on the day before infection in T175 flasks. TRES6 and TRES328 antibodies were incubated for one hour at a concentration of 200 ng/ml with 2x10 6 TCID50 of the CoV2-ER1 virus. Subsequently, the medium in the cell culture flasks was exchanged for 20 ml of fresh medium and the virus/antibody mix was added to the cell cultures. Control cultures were inoculated with 2x10 6 TCID50 of the CoV2-ER1 virus in the absence of any antibody. The cells were checked daily for infection. If infection occurred the CT value was assessed by real time PCR. In case of a break through infection 100 µl of the break through viral supernatant were incubated with the antibody from the previous round at twice its concentration and added onto fresh Vero-E6 cells. This procedure was repeated five times. The TCID50 of the passaged virus stocks were measured and the neutralization against both TRES antibodies was assessed. Additionally the viral RNA was isolated from cell culture supernatant with a PureLink RNA Mini Kit (Thermo Fisher Scientific, Waltham, USA). The viral RNA was then sequenced for the identification of the escape mutations. Briefly, commercial reagents were used for unique dual indexed amplicon library generation according to the Artic protocol (NEBNext® ARTIC SARS-CoV-2 Library Prep Kit (Illumina®)), which were sequenced with MiSeq® Reagent Kit v2 (500 Cycles) on the Illumina MiSeq instrument and data analyzed with CLC Genomic Workbench 21 (Qiagen, Aarhus, Denmark). Median coverage obained for SARS-CoV-2 P5 1914, for SARS-CoV-2 TRES328 1326 and for SARS-CoV-2 TRES6 1436.

Animal experiments were performed following the EU Directive 2010/63/EU for animal experiments and were approved by local authorities after review by an ethical commission (TVV 21/20). Thirty-six female K18-hACE2 mice (Jackson Laboratory, Bar Harbor, USA) were infected intranasally under isoflurane anesthesia with 300 FFU of SARS-CoV-2 strain MUC-IMB-1 p.1 in a total volume of 50 µl. 24 hours after virus inoculation, twelve mice per group were injected intravenously either with 5.25 mg/kg body weight of TRES6 or TRES328 antibody or an isotype control antibody (TRES480) in a total volume of 100 µl. The animals were scored daily and the survival and disease incidence were measured over a maximal 10-day period. One cohort (n=6) was euthanized at day 4, while the second cohort (n=6) was scored for humane endpoints and euthanized at day 10 post virus inoculation. After euthanasia, BALs (bronchoalveolar lavages) were collected, and lung, brain, liver and spleen were taken. The organs were homogenized in 2 ml PBS using gentleMACS M tubes and gentleMACS Octo Dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany). Afterward, the tubes were centrifuged at 2000 x g for 5 min at 4°C to discard cell debris. Viral RNA was isolated from 140 µl of homogenated supernatants or BAL using QIAamp Viral RNA Mini Kit (Qiagen, Aarhus, Denmark). The viral load in indicated organs and BAL was analyzed by RT-qPCR [64] . Reactions were performed using TaqMan® Fast Virus 1-Step Master Mix (Thermo Fisher Scientific, Waltham, USA) and 5 µl of isolated RNA as a template. Synthetic SARS-CoV-2 RNA (Twist Bioscience, San Francisco, USA) was used as a quantitative standard to obtain viral copy numbers. The viral load reduction was calculated in comparison to the isotype control. The clinical scoring system included the following items: weight loss and body posture (0-20 points), general conditions including the appearance of fur and eye closure (0-20 points), reduced activity and general behavior changes (0-20 points), and limb paralysis (0-20 points). Mice were euthanized when reaching a cumulative score of 20.

For the detection of infectious virus in BAL Vero-E6 cells were seeded at 2x10 4 cells/well in a 96-well plate in 200 µl of DMEM, 10 % FCS, 1x penicillin/streptomycin (Thermo Fisher Scientific, Waltham, USA) 24 h before infection with 20 -100 µl of BAL diluted in DMEM with 1x penicillin/streptomycin for 3 hours. After replacing the supernatant with overlay medium (DMEM with 1 % methylcellulose, 2 % FCS and 1x penicillin/streptomycin), cells were incubated for 27 hours. SARS-CoV-2 infected cells were visualized using SARS-CoV-2 S protein specific immunochemistry staining with anti-SARS-CoV-2 spike glycoprotein S1 antibody (Abcam, Cambridge, Great Britain) as described previously [64] . Statistical evaluation of the data was performed by Kruskal-Wallis test (one-way ANOVA) and Dunn's Pairwise Multiple Comparison Procedures as post hoc test.

A Novel Coronavirus from Patients with Pneumonia in China

COVID-19 vaccines: where we stand and challenges ahead

SARS-CoV-2 vaccines strategies: a comprehensive review of phase 3 candidates. npj Vaccines

Estimating real-world COVID-19 vaccine effectiveness in Israel using aggregated counts

Early effectiveness of COVID-19 vaccination with BNT162b2 mRNA vaccine and ChAdOx1 adenovirus vector vaccine on symptomatic disease, hospitalisations and mortality in older adults

Potent SARS-CoV-2 Neutralizing Antibodies Directed Against Spike N-Terminal Domain Target a Single Supersite

A neutralizing human antibody binds to the N-terminal domain of the Spike protein of SARS-CoV-2

Structure-based development of human antibody cocktails against SARS-CoV-2

Potent neutralizing antibodies from COVID-19 patients define multiple targets of vulnerability

Studies in humanized mice and convalescent humans yield a SARS-CoV-2 antibody cocktail

Prevalent, protective, and convergent IgG recognition of SARS-CoV-2 non-RBD spike epitopes in COVID-19 convalescent plasma

Potent neutralizing antibodies against multiple epitopes on SARS-CoV-2 spike

Potent Neutralizing Antibodies against SARS-CoV-2 Identified by High-Throughput Single-Cell Sequencing of Convalescent Patients' B Cells

SARS-CoV-2 501Y.V2 escapes neutralization by South African COVID-19 donor plasma

Human monoclonal antibodies block the binding of SARS-CoV-2 spike protein to angiotensin converting enzyme 2 receptor

Human neutralizing antibodies elicited by SARS-CoV-2 infection

Convergent antibody responses to SARS-CoV-2 in convalescent individuals

Potently neutralizing human antibodies that block SARS-CoV-2 receptor binding and protect animals. bioRxiv : the preprint server for biology

Monoclonal Antibodies for Prevention of SARS CoV-2 Infection Asymptomatic in Healthy Adults and Adolescents Who Are Household Contacts to an Individual With a Positive SARS-CoV-2 RT-PCR Assay, Identifier: NCT04452318. 2020

ClinicalTrials.gov. Safety, Tolerability, and Efficacy of Anti-Spike (S) SARS-CoV-2 Monoclonal Antibodies for the Treatment of Ambulatory Adult Patients With COVID-19

An update to monoclonal antibody as therapeutic option against COVID-19

SARS-CoV-2 D614G variant exhibits efficient replication ex vivo and transmission in vivo

Receptor utilization of angiotensin-converting enzyme 2 (ACE2) indicates a narrower host range of SARS-CoV-2 than that of SARS-CoV

SARS-CoV-2 B.1.1.7 and B.1.351 Spike variants bind human ACE2 with increased affinity. bioRxiv : the preprint server for biology

Resurgence of COVID-19 in Manaus, Brazil, despite high seroprevalence

Comparative infectivity and pathogenesis of emerging SARS-CoV-2 variants in Syrian hamsters

High-Throughput, Single-Copy Sequencing Reveals SARS-CoV

Spike Variants Coincident with Mounting Humoral Immunity during Acute COVID-19. bioRxiv : the preprint server for biology

mRNA vaccine-elicited antibodies to SARS-CoV-2 and circulating variants

Serum Neutralizing Activity Elicited by mRNA-1273 Vaccine

Neutralization of SARS-CoV-2 spike 69/70 deletion, E484K and N501Y variants by BNT162b2 vaccine-elicited sera

Resistance of SARS-CoV-2 variants to neutralization by monoclonal and serumderived polyclonal antibodies

Decreased neutralization of SARS-CoV-2 global variants by therapeutic anti-spike protein monoclonal antibodies

A new coronavirus associated with human respiratory disease in China

Structural and functional properties of SARS-CoV-2 spike protein: potential antivirus drug development for COVID-19

Structure, Function, and Evolution of Coronavirus Spike Proteins

Exosomal vaccines containing the S protein of the SARS coronavirus induce high levels of neutralizing antibodies. Virology

Investigation of a COVID-19 outbreak in Germany resulting from a single travel-associated primary case: a case series. The Lancet. Infectious diseases

Virological assessment of hospitalized patients with COVID-2019

Tracking Changes in SARS-CoV-2 Spike: Evidence that D614G Increases Infectivity of the COVID-19 Virus

Molecular Architecture of Early Dissemination and Massive Second Wave of the SARS-CoV-2 Virus in a Major Metropolitan Area. mBio

Physical map of the mouse λ light chain and related loci

Identification of Common Deletions in the Spike Protein of Severe Acute Respiratory Syndrome Coronavirus 2

A Single-Dose Intranasal ChAd Vaccine Protects Upper and Lower Respiratory Tracts against SARS-CoV-2

Antibodydependent enhancement and SARS-CoV-2 vaccines and therapies

Longitudinal Isolation of Potent Near-Germline SARS-CoV-2-Neutralizing Antibodies from COVID-19 Patients

A human monoclonal antibody blocking SARS-CoV-2 infection

Focused Evolution of HIV-1 Neutralizing Antibodies Revealed by Structures and Deep Sequencing

Broad neutralization coverage of HIV by multiple highly potent antibodies

Somatic mutations of the immunoglobulin framework are generally required for broad and potent HIV-1 neutralization

Efficient generation of monoclonal antibodies from single human B cells by single cell RT-PCR and expression vector cloning

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

CD4(+) T Cells Induced by Tuberculosis Subunit Vaccine H1 Can Improve the HIV-1 Env Humoral Response by Intrastructural Help. Vaccines

Monoclonal antibodies to discriminate the EF hand containing calcium binding adaptor proteins EFhd1 and EFhd2. Monoclonal antibodies in immunodiagnosis and immunotherapy

Rapid response flow cytometric assay for the detection of antibody responses to SARS-CoV-2

Modulation of Vaccine-Induced HIV-1-Specific Immune Responses by Co-Electroporation of PD-L1 Encoding DNA. Vaccines

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

A highly specific and sensitive serological assay detects SARS-CoV-2 antibody levels in COVID-19 patients that correlate with neutralization

No evidence for the use of DIR, D-D fusions, chromosome 15 open reading frames or VH replacement in the peripheral repertoire was found on application of an improved algorithm, JointML, to 6329 human immunoglobulin H rearrangements

IMGT/V-QUEST: the highly customized and integrated system for IG and TR standardized V-J and V-D-J sequence analysis

Bamlanivimab does not neutralize two SARS-CoV-2 variants carrying E484K in vitro

Quantitative Assays for Virus Infectivity

Growth, detection, quantification, and inactivation of SARS-CoV-2. Virology

We would like to thank Isabell Schulz and Doris Jungnickl for her excellent technical assistance. We kindly thank Jasmin Fertey and Rosina Ehmann for providing high titer virus stocks of SARS-CoV-2 and Alexandra Rockstroh for the optimized SARS-CoV-2 detection protocol. The expression plasmid for the spike ectodomain was kindly provided by Jason McLellan, Austin, USA, the TriGrid electrode array for DNA electroporation by Drew Hannaman ,Ichor Medical Systems, Inc. and S2 and S1 protein by Thomas Schumacher, Virion/Serion GmbH, Würzburg. The ELISPOT Analyzer was obtained with financial support from Fondation Dormeur, Vaduz. This work was supported by a grant (01KI2043) from the Bundesmininisterium für Bildung und Forschung (BMBF) and funds from the Bavarian State minstery for Science and the Arts to K.Ü., H-M J., and T.H. W. Further support was provided by B-FAST and COVIM, two BMBF-funded projects of the Netzwerk Universitätsmedizin (NaFoUniMedCovid19; FKZ: 01KX2021) and a DFG-funded research training group (RTG 2504).The authors declare that they complied with all relevant ethical regulations. All relevant data supporting the findings of this study are available within the paper and its supplementary information files.