key: cord-0955221-9uvwagos
authors: Plaper, Tjaša; Aupič, Jana; Dekleva, Petra; Lapenta, Fabio; Manček Keber, Mateja; Jerala, Roman; Benčina, Mojca
title: Coiled-coil heterodimers with increased stability for cellular regulation and sensing SARS-CoV-2 spike protein-mediated cell fusion
date: 2020-12-10
journal: bioRxiv
DOI: 10.1101/2020.12.10.419440
sha: 69bb01a2a5316666bae8a81088d56c5d7a7f5bef
doc_id: 955221
cord_uid: 9uvwagos

Coiled-coil (CC) dimer-forming peptides are attractive designable modules for mediating protein association. Highly stable CCs are desired for biological activity regulation and assay. Here, we report the design and versatile applications of orthogonal CC dimer-forming peptides with a dissociation constant in the low nanomolar range. In vitro stability and specificity was confirmed in mammalian cells by enzyme reconstitution, transcriptional activation using a combination of DNA-binding and a transcriptional activation domain, and cellular-enzyme-activity regulation based on externally-added peptides. In addition to cellular regulation, coiled-coil-mediated reporter reconstitution was used for the detection of cell fusion mediated by the interaction between the spike protein of pandemic SARS-CoV2 and the ACE2 receptor. This assay can be used to investigate the mechanism and screen inhibition of viral spike protein-mediated fusion under the biosafety level 1conditions.

The coiled-coil (CC) motif, ubiquitous in natural proteins, is characterized by the relatively well-defined assembly 2 and specificity rules and is highly attractive as a building block for designing and developing novel polypeptide 3 scaffolds 1 , functional proteins 2-5 , and as a regulator of protein-protein interactions 6-8 . CC protein origami (CCPO) 4 cages 9,10 , polymer-peptide hydrogels, nanoparticles, and peptide fibers 11 are fine demonstrations of the utility of 5

CCs for engineering new polypeptide nanostructures. The specificity of CC pairs has been harnessed in order to 6 develop logic circuits and controllers, as well as mammalian cell engineering 4,12 , however different applications, 7 require different CC stabilities [13] [14] [15] [16] [17] [18] . 8 9

Since the stabilities of several previously-designed four-heptad long CC pairs either had low orthogonality to other 10 proteins (e.g., E/K pair 13 ) or were only moderately stable 16, 19 , we sought to design a set of highly stable orthogonal 11 parallel heterodimeric CCs that could improve the sensitivity and increase the range of existing or prospective 12 biological applications. 13

The COVID-19 pandemic caused by SARS-CoV-2 infection is currently among the most acute problems of 15 humanity 20 , and investigation of its function and discovery of inhibitors command the attention of many scientists. 16

The enveloped SARS-CoV-2 viruses require fusion of viral and cellular membranes in order to infect host cells 21 . 17

The membrane fusion process is assisted by viral fusion proteins located at the viral envelope. After binding of 18 viral fusion proteins to the receptor at the host cell membrane, a structural rearrangement of the viral fusion protein 19 lowers the energetic barrier of the coalescence of the two biological membranes and the introduction of the viral 20 genome into the cytosol occurs 22 . Multiple methods are used for the evaluation of the viral entry and alternative 21 steps of this process 23 , from active virus to pseudotyped virus (a.k.a. pseudovirus) infection, which require BSL-22 2 or BSL-3 labs in order to prevent the danger of infection. Alternatively, cell-cell fusion assays can be used, 23 based on the formation of syncytia between cells that express viral fusion (spike) proteins and the cellular receptors 24 (ACE2) on other cells, offering a safer, virus-free alternative for viral entry studies and the development of antiviral 25 drugs. Syncytium formation has been traditionally analyzed by slow and semi-quantifiable microscopy [24] [25] [26] . 26

Generation of a signal, which occurs only after the fusion between donor and acceptor cells 27 , provides a good 27 basis for this sensitive method to monitor syncytium formation. Therefore, split fluorescence or luminescence 28 proteins linked to synthetic dimerization domains that form very stable dimers would be a prime choice as 29 reporters. 30 31 In order to generate a set of CC heterodimers with the high thermodynamic stability necessary for new biological 32 applications, we built on a previously published orthogonal NICP CC set 17 but increased the stability of the peptide 33 pairs by reducing the number of asparagine (Asn, N) residues 28,29 and modifying amino-acid residues at the 34 noncontact b, c, and f positions 16 . CC pairs were able to reconstitute a split reporter enzyme, and we additionally 35 demonstrate that their activity could be regulated externally by the addition of peptides to cells. Similar to the split 36

reporter reconstitution, we also demonstrated reconstitution of a CC split transcription factor. Moreover, we sought 1 to employ the designed CC pairs in monitoring membrane fusion based on the interaction between SARS-CoV-2 2 spike protein and ACE2 receptor interactions. This sensitive syncytium formation assay can be used to monitor 3 the fusion process and screen for inhibitors in a safe and easy-to-employ way. 4 5

De novo designed orthogonal heterodimeric CC pairs with high stability 7

Using previously established principles 16,17 , we designed four heptad long CC dimer-forming peptides (named 8 N5-N8) expected to form two orthogonal CC pairs in a parallel orientation with a high binding affinity ( Table 1 , 9 N-variants; a CC N7:N8 structural model is depicted in Fig S1A) . While Asparagine-Asparagine (Asn-Asn) 10 pairing at the a positions is important for ensuring interaction specificity and peptide orientation 13 , it contributes 11 unfavorably to the affinity 28,30 . Therefore, in order to increase the affinity of the NICP set, we maintained only a 12 single Asn-Asn interaction per CC pair. The Asn (N) residue was placed at the a position in the third heptad for 13 peptides N5 and N6 and in the second heptad for peptides N7 and N8 to ensure their mutual orthogonality. All 14 other a positions were occupied by isoleucine (Ile, I), whereas leucine (Leu, L) was placed at the d position in all 15 heptad repeats. Given that the residues at the surface-exposed positions do not contribute to the pairing interactions 16 16 , alanine (Ala, A) residues were placed at the surface-exposed b, c, and f positions, in order to promote high 17 helical propensity and, consequently, increase CC stability. In order to ensure orthogonality, lysine (Lys, K) and 18 glutamic acid (Glu, E) were positioned at the e and g positions, obtaining peptide pairs with a unique electrostatic 19 motif that should promote the formation of two CC heterodimers named N5:N6 and N7:N8-both in parallel 20 orientations. The electrostatic motifs of these N-type peptides were kept similar to those of the previously 21 published NICP P-type peptides (P5-P8) 17 (Table 1, Table S1 ). The interaction pattern of the designed set was 22 evaluated by the scoring function 31 , supporting the favorable formation of N5:N6 and N7:N8 CCs (Fig S1B) 

Following peptide design, thermal stability and orthogonality of synthetic N-type peptides were analyzed. The 1 secondary structure and stability of peptides were probed by circular dichroism (CD) spectroscopy. The designed 2 peptide pairs, N5:N6 and N7:N8, were orthogonal to each other (Fig 1E) , with the characteristic α-helical CD 3 spectra, Tm above 70 ºC (Fig 1A) , and a dissociation constant determined by the ITC (Kd ITC ) in the low nanomolar 4 range (1-15 nM; Fig 1B) . In addition, size-exclusion chromatography coupled to static light scattering (SEC-5 MALS) analysis showed that the equimolar mixture of peptides preferentially formed heterodimers (Fig S2B) ; 6 however, all four individual N-peptides were α-helical at 20 ºC (Fig S2A) . The melting curves of peptides were 7 sigmoidal with a Tm midpoint between 37 ºC and 55 ºC (Fig 1A) . Analysis by SEC-MALS confirmed the 8 homodimeric formation of all peptides in solution (Fig S2B) . 9

10 In order to tune the stability of designed CCs and avoid homodimerization, we designed additional PA peptide 11 pairs. The Ile residues at the a positions of the first or fourth heptad were replaced with Asn, resulting in peptide 12 variants named P5A:P6A and P7A:P8A (PA peptides), respectively ( Table 1 ; the structural model of P7A:P8A is 13 depicted in Fig S1A) . While the inclusion of an additional Asn residue destabilized the hydrophobic core 28-30 , it 14 also led to a decrease in the predicted helicity; thus, negatively contributing to the CC formation as confirmed 15 experimentally. The CD spectra and thermal denaturation scans of the individual peptides showed that the second 16

Asn within the PA peptide decreased their α-helical and thermal stability (Fig S2C) . Furthermore, in contrast to 17 the N variants, the individual PA peptides were present in solution in a monomeric state and formed dimers only 18 when mixed with their cognate peptide, as shown by SEC-MALS analysis (Fig S2D) . The Tm values of the 19 P7A:P8A and P5A:P6A CC pairs were 64 ºC and 57 ºC, respectively (Fig 1C) . The ITC-measured Kd value was 20 in the nanomolar range (Fig 1D) . The orthogonality of PA peptides was confirmed by measuring the stability of 21 all combinations of PA and N-type peptides (Fig S2E) . The expected cognate peptide pairs exhibited higher Tm 22 values than any other combination of peptides, indicating the preference of the peptides to bind to their designated 23 partners. 24 25

Characterization of CC pairs in vitro provided essential information about their stability. Next, the impact of a 27 cellular environment on the formation of CCs between the cognate peptides was tested by the reconstitution of a 28 split luciferase reporter in mammalian cells (scheme Fig 1F) . The N-and C-terminal segments of split luciferase 29 (henceforth, nLuc and cLuc, respectively) were genetically tethered to the N-and C-termini of the N8 and N7 30 peptides, respectively. All four combinations of CC-split luciferase reconstituted enzymatic activity in HEK293 31 cells (Fig 1G, Fig S3A) . The highest activity was measured for the CC-split luciferases linked to the N-termini of 32 cognate CC peptides; therefore, this setup was used for further experiments. 33 34 After establishing the most efficient setup of the CC-split luciferase, we analyzed the efficiency of CC formation 35 for newly designed N-peptides (N5:N6 and N7:N8) and compared those with the PA-peptides (P5A:P6A and 36 P7A:P8A; Fig 1H, Fig 1I) . Both the N-and PA-peptide pairs reconstituted the CC-split luciferase; however, 1 reconstitution of the PA-CC-split luciferases was less effective than the N-CC-split luciferase, which corroborated 2 the in vitro stability results of the CC peptides' stability, regardless of the homodimerization propensity of the N 3 peptides. The orthogonality of all eight peptides was also tested in mammalian cells. In accordance with in vitro 4 results, the nLuc:N6 paired only with cLuc:N5 or P5A but not with cLuc:N7 or P7A, and nLuc:N8 paired with 5 cLuc:N7 and P7A (Fig 1J) . The reconstitution of N-CC-split luciferase was more efficient compared to PA-CC-6 split luciferase, as reflected by the in vitro thermal stability of N-and PA-CCs, which makes N-type peptides very 7 suitable for biological applications. While CD-based thermal denaturation and SEC-MALS cannot discriminate 8 between the homo-and hetero-dimeric CCs, the reconstitution of CC-split proteins depends on the formation of 9 CC heterodimers 3 and, therefore, confirms the heterodimerization of CC pairs. 10 pairs. The binding isotherms of heat release per injection are depicted as a function of the increasing peptide-to-6 peptide molar ratio. The dissociation constant, Kd ITC , was calculated using the two-state dimer association model. Luciferase activity of reconstituted CC-split luciferase in HEK293T measured 48 h after transfection of HEK293T 10 cells with a plasmid expressing a combination of nLuc tethered to N8 and cLuc tethered to N7 peptide. (H,I) 11

Luciferase activity determined 48 h after transformation of HEK293T cells with plasmids expressing nLuc:N8 or 12 nluc:N6; and cLuc tethered to N7 or P7A or cLuc tethered to N5 and P5A. (J) Orthogonality of designed N peptide 13 set in HEK293T cells by co-cotransfection of the nLuc:N8 or nLuc:N6 fusion encoding plasmid, and cLuc:CC 14 (CC stands for N5, P5A, N7, P7A). Reconstituted luciferase activity was measured 48 h after transfection. 15

Amounts of plasmids are indicated in Table S2 . The values (G-J) represent the means (± s.d.) from four 1 independent cell cultures, individually transfected with the same mixture of plasmids, and are representative of 2 two independent experiments. 3 4

We have previously designed and reported an orthogonal, parallel NICP set of CC heterodimers 17 that contain the 6 same electrostatic motifs as the N-and PA-peptides. Pairing CC-forming peptides with complementary 7 electrostatic motifs (EKEK:KEKE and EKKE:KEEK at e and g positions) but different helical propensities at non-8 interacting b, c, and f heptad positions allows for facile fine-tuning of the stability of the resulting CC combinations 9 16 . Consequently, pairing N-peptide variants with P variants (Table S1 sequences) can create an extended range 10 of pairs with different stabilities. Using the CC-split luciferase assay described above (Fig 1F) , we analyzed the 11 formation of N-peptides in combination with the P-peptides with a complementary electrostatic motif. NLuc was 12 genetically tethered to the N-termini of the N8 or N6 peptide, and cLuc to N7, P7A, P7, P7SN, N5, P5A, P5, or 13 P5SN peptide (Fig 2A, Fig 2B) . The interaction matrix of all electrostatically complementary peptide 14 combinations predicted the formation of CCs with a range of stability profiles, which were validated 15 experimentally by thermal denaturation scans (Fig S4A, Fig S4B) . The stability of CCs formed with the cognate 16 N-and P-peptide pairs (Tm 43-68 ºC) was lower compared to the N-type (Tm > 70 ºC). We also investigated 17 whether the stability of the cognate CC peptides with different helical propensities, determined in vitro, correlated 18 with the reconstitution efficacy of CC-split luciferase in mammalian cells. The highest luciferase activity was 19 determined for the combination of nLuc:N8 with cLuc:N7 and nLuc:N6 with cLuc:N5 (Fig 2A, Fig 2B; for CC-20 split luciferase see Fig S3B) . In contrast, the CC-split luciferase combination, nLuc:N with cLuc:PSN pairs, 21 generated the weakest signal. The results indicate that CC-split luciferase, as a reporter of CC stability in cells, 22 adequately reflects the CC stability determined in vitro. As predicted, only orthogonal peptide variants with 23 complementary electrostatic motifs reconstituted the functional luciferase (Fig 2C, Fig 2D) . Furthermore, the wide 24 range of CC stabilities obtained for all the pairs enables the use of different CC-dimerizing elements to fine tune 25 the enzymatic activity. experiments. Amounts of used plasmids are indicated in Table S2 . 8 9

Regulation of split luciferase activity in cells by a displacer peptide 10

The stability of CCs and their formation in mammalian cells can be tuned by combining CC-forming peptide pairs 11 with different helical propensities. Thus, we examined whether the electrostatically cognate CC-forming peptide 12 (CC-displacer peptide), with a higher helical propensity, could displace the weaker interacting peptide within CC 13 in mammalian cells (scheme Fig 3A) . CC-split luciferase was used as a reporter, and non-labeled peptides were 14 used as competitors. Initially, we tested whether co-expressed CC-displacer peptide, N7, could compete with CC 15 peptides tethered to the split luciferase (Fig 3B) . Co-expression of the N7 CC-displacer peptide attenuated 16 reconstitution of the split luciferase tethered to peptides N8:P7SN, N8:P7, or N8:P7A (displacer peptide N5 17 attenuated N6:P5 split luciferase activity; Fig S5A) but, as expected, failed to effectively compete with the 18 immensely stable, N8:N7 CC (Fig 3C) . 19 20 Since the externally added peptide provides a more direct and potentially therapeutically relevant way to 21 manipulate the activity of split proteins than co-expression of the CC-displacer peptide, we tested the efficacy of 22 strand displacement by adding synthetic CC-displacer peptide to cells expressing reconstituted CC-split luciferase 23 (Fig 3D) . Three CC-displacer peptides with a cognate electrostatic motif (N7, P7A, and P7) were tested on four 24 combinations of CC-split luciferase: N8:N7, N8:P7A, N8:P7, and N8:P7SN (Fig 3E, Fig S5B) . As predicted, N7 25 displacer peptide, with its high-helical propensity, debilitated CC-split luciferase assembly most efficiently only 2 1 hrs after the addition of a synthetic displacer peptide (Fig 3E) . The P7A, P7, and P7SN peptides, for which the 2 midpoint Tm of N-P-CCs was more than 10 ºC lower than N-CCs, failed to displace the cognate peptide from the 3 N8:N7 split luciferase as efficiently (Fig S5B) . Additionally, the N5 peptide effectively displaced cLuc:P5 from 4 the N6:P5 split luciferase (Fig S5C) . 5 6 Moreover, the CC-displacer peptide with mismatched electrostatic motif failed to displace the cognate CC peptide 7 pair and, therefore, failed to attenuate the luciferase activity (Fig 3G, Fig S5D) . Taken together, the results indicate 8 that the formation and disruption of CC-mediated pairs can be regulated by competition with an electrostatically 9 cognate peptide that forms stronger interactions. Initially, we tested a reconstitution of TALE-A:N8 or TALE-A:N6 with CC:VP16 (Fig 4B, Fig S6A) . The highest 20 expression of reporter luciferase under TALE promoter was detected for the combination of TALE-A:N8 with 21 N7:VP16 and TALE-A:N6 with N5:VP16 (Fig 4B) . As expected, VP16 tethered to a PA peptide variant with 22 lower helical propensity resulted in a weaker transcription of the reporter. The efficiency of the reconstituted 23 transcription factor was more dependent on the amount of the transcriptional activation-domain-(CC:VP16)-24 encoding plasmid than on DNA binding domain (TALE-A:CC), which can saturate the binding sites (Fig 4B, Fig  25   S6A ). Only the cognate CC peptide pairs (TALE-A:N6 with N5:VP16 and TALE-A:N8 with N7:VP16) triggered 26 transcription of the reporter (Fig S6B) . The results show that CC peptides with different stabilities can mediate the 27 efficiency of transcriptional activation, with expression reflecting the stability of the CC pairs. 28 29 Orthogonal CC peptides provide a very effective means of constructing multifunctional proteins. In addition to 30 genetically fused TALE-A to a single CC-peptide, we also created a construct, where the DNA binding domain 31 was fused to two consecutive orthogonal CC-peptides (N6 and N8; scheme Fig 4C) . In order to demonstrate the 32 usability of orthogonal peptides, the TALE-A:N6:N8 was co-expressed with the CC:VP16 activator and/or 33 CC:KRAB repressor domain. While the combination of TALE-A:CC with CC:VP16 acted as a transcriptional 34

activator, the addition of CC:KRAB suppressed transcription (Fig 4C) . Addition of the KRAB repressor tethered 35 to CC N5, P5A, P5, or P5SN peptides (CC:KRAB), with different affinities to N6 of the reconstituted TALE-36 A:N6:N8 with N7:VP16, effectively suppressed transcription, correspondingly with their CC stability (Fig 4D) . A 37 relatively low amount of co-transfected CC:KRAB plasmid was required for the effective suppression of the 38 transcription activator. Even when VP16 was replaced with the VPR activation domain, which is regarded as a 1 stronger domain 39 , the co-expressed CC:KRAB domain nevertheless suppressed the transcription (scheme Fig  2   S6D, Fig S6E) . The stability of CC formed between the CC:KRAB peptide and TALE-A:N6 determined the 3 degree of transcription suppression, with P5SN:KRAB being the weakest suppressor. Therefore, the same DNA 4 binding protein could be active, whether up-or down-regulated, which also represents a Boolean logic function 5 B, NIMPLY A. 6 7

The activity of the multifunctional transcription factor, TALE-A:CC, could be regulated by CC:VP16 and 8 CC:KRAB association as well as by the addition of a CC-displacer peptide. The same strategy of the CC-displacer 9 peptide used for CC-split luciferase (Fig 3) could be used to regulate the repressor activity of TALE:N6:N8 with 10 N7:VP16 (scheme Fig 4E) and P5:KRAB (scheme Fig 4G) . The reconstitution of the CC-split transcription factor, 11 similar to that of CC-split luciferase, was attenuated by co-expressed CC-displacer peptide tagged with a nuclear 12 localization signal (NLS). The degree of attenuation reflected the stability of CC within the CC-transcription factor 13 and CC-displacer peptide (Fig 4F, Fig S6C) . As predicted, the N7 displacer peptide weakened the transcription 14 activator assembly most efficiently. P7A and P7, with a weaker affinity toward N8, failed to prevent the 15 reconstitution of a functional transcription as efficiently (Fig S6D) . Besides, through the addition of increasing 16 amounts of N5 NLS peptide (cognate to N6), we were able to displace the KRAB suppressor domain and regain 17 transcriptional activity (Fig 4G, 4H) . Altogether, we demonstrated that the array of orthogonal CC peptide pairs 18 that form CCs with various ranges of stability represents an effective tool for rapid and effective regulation of 19 biological activity, likely transferrable to other enzymes, assemblies, and processes. 20 two independent experiments. For amounts of plasmids, see Table S2 . Statistical analyses and the corresponding 19

p-values are listed in Table S3 . 20 21

We realized that the high affinity between CC domains could be used to monitor the fusion of cells for the 2 syncytium formation. SARS-CoV-2 entry into host cells is mediated by a viral spike (S) protein, which binds to 3 the human Angiotensin-converting enzyme 2 (ACE2) as a viral entry receptor 40 . Viral entry depends on the 4 binding of the S1 domain of spike protein to the ACE2 receptor, which facilitates viral attachment to the surface 5 of target cells. Virus entry also requires the processing of the spike protein by the host transmembrane serine 6 protease 2 (TMPRSS2), which exposes the S fusion peptide. This process enables the fusion of the viral and 7 cellular membranes, a process driven by the spike protein's S2 subunit 24,41 . Monitoring membrane fusion triggered 8 by the binding of a viral fusion protein to the receptor represents an important assay for monitoring this key process 9 in viral infections. Many inhibitors, as well as vaccines, target viral attachment to host cells and the fusion process. 10 Experiments using active viruses or pseudoviruses are labor-intensive, slow, and require specific safety measures. 11

Fluorescent proteins and transcriptional activation in target cells have been used to monitor fusion triggered by the 12 S protein of SARS-CoV-2 with ACE2 receptor 40 . An assay was prepared based on the fusion of cells expressing 13

of cell fusion was detected by co-transfection of two cell groups with split luciferase reporters (cLuc:P7A and 15 nLuc:P8A or cLuc:N7 and nLuc:N8) with ACE2 and SARS-CoV-2 S protein, respectively (Fig 5A) . Mixing of 16 donor and acceptor cells resulted in a rapid reconstitution of the reporter in fused cells. The addition of CCs to a 17 split luciferase reporter acts as a strong dimerization domain, generating a high signal following cell fusion, but 18 only for cell combinations that expressed S protein and hACE2 (Fig 5B) . 19 20

In addition to a CC-split luciferase reporter, we set out to design a CC-split green fluorescent protein (GFP) 21 reporter 42 , which enables syncytium visualization via confocal microscopy and quantification using flow 22 cytometry. The fusion of CC peptides to the split GFP improves the efficiency of the reporter reconstitution and 23 enables the detection of syncytia. We designed GFP1-10 genetically fused to the N7 or P7 peptide (GFP1-10:N7, 24 GFP1-10:P7) and GFP11 fused to N8 or P8 peptide (N8:GFP11, P8:GFP11; Fig 5C) . The donor cells expressed S 25 protein, iRFP or mCherry, and CC-split GFP (GFP1-10:N7 or GFP1-10:P7). The acceptor cells expressed ACE2 26 receptor, BFP, and CC-split protein. Using CC-split GFP, we were able to detect cell-cell fusion both utilizing a 27 confocal microscope and via flow cytometry (Fig 5D, 5E) . All syncytia also expressed iRFP and BFP. In order to 28 increase the sensitivity of flow cytometry for syncytium detection, a construct with linked three repeats of 29 N8:GFP11 (3×(N8:GFP11)) was designed (Fig 5D) , which enables three molecules of GFP to reconstitute in a 30 fluorescent reporter. Indeed, this kind of reporter proved to be even more effective compared to the construct 31 containing only a single CC-forming peptide (Fig 5E) . 32

The S protein-/ACE-mediated fusion process could, thus, be monitored at different stages, and inhibitor 34 efficiencies of this process could be detected. We demonstrated the effect of inhibiting TMPRSS2 protease that is 35 required to cleave S protein and expose the fusion protein of spike protein prior to fusion. Camostat mesylate 43 36 has been previously tested as an inhibitor in the active viral assay 43 (Fig 6A) . Cell fusion was detectable without 1 the presence of TMPRSS2 protease (Fig S7A) ; however, a strong signal increase was observed upon the co-2 expression of the protease (Fig. 6B) , which confirms the role of TMPRSS2 in CoV-2 spike-protein-mediated 3 syncytia formation as previously reported 25 . Additionally, fusion could be inhibited by blocking the ACE2 4 receptor by the addition of a soluble RBD domain 44,45 (Fig.6C, 6D) . Increasing amounts of RBD protein indeed 5 reduced the number of syncytium cells generated by restraining ACE2 and CoV-2 S protein interaction (Fig 6D) . 6

Inhibition of cell-cell fusion was also detected using flow cytometry. shown. For amounts of plasmids, see Table S2 . Statistical analyses and the corresponding p-values are listed in 30  Table S3 . Table S2 . 14 Statistical analyses and the corresponding p-values are listed in Table S3 .

In this study, we designed and explored the potential of CC-peptide pairs with high stability as a tool for biological 18 applications. We expanded the validated orthogonal set of CC dimer-forming peptides 9 with two de novo designed Finally, high-affinity CC pairs were used to set up the assay for monitoring cell fusion. Genetic fusion to protein 13 interaction domains can guide protein complex formation 46 , and strong affinities are needed for a robust output in 14 comparison to relying solely on the reconstitution of split GFP 43 . We established an assay to monitor the SARS-15

CoV-2 spike-protein-mediated cell-cell fusion, which allowed for quantitation by reconstitution of a CC-split 16 luciferase reporter or visualization of syncytia by the reconstituted CC-split GFP. The split luciferase reporter 17 genetically fused to CCs produced a robust response upon syncytia formation, guided by the interaction between 18 SARS-CoV-2 spike protein and ACE2 receptor. Addition of the TMPRSS2 inhibitor, Camostat, or RBD, which 19 targets different steps in the fusion process, inhibited syncytia formation, validating the proteinase in S protein 20 processing and binding to ACE2 receptor as targets for viral entry. The strategy of self-reconstituted split GFP 21 previously described by Kamiyama et al. 42 was used to further intensify the generated fluorescent signal so it 22 could detect formed syncytia upon mixing donor and acceptor cells, enabling better visualization and flow 23 cytometric analysis. This rapid (~ 3hrs), sensitive assay enables screening of the inhibitors of this process as 24 potential drugs to prevent viral infection. In conclusion, high-affinity CC dimers represent valuable tools for 25 inquiring and regulating many biological processes. 26 27

Peptides, plasmids, and cell lines. 2

The peptides used in this study (Proteogenix, France) were protected at the N-and C-termini by acetylation and 3 amidation, respectively. All peptides were dissolved in a stock concentration of approximately 1 mg/ml in 4 deionized water, except for the water-insoluble N1 peptide, which was dissolved in 10 % (w/v) ammonium 5 bicarbonate. The exact peptide concentrations were determined based on absorbance at 280 nm, and the extinction 6 coefficients were calculated by the ProtParam web tool (https://web.expasy.org/cgi-bin/protparam/protparam). All 7 plasmids (listed in Table S4 ) were constructed using the Gibson assembly method. A human embryonic kidney 8 cell line, HEK293T (ATCC CRL-3216), was cultured in complete media (DMEM; 1 g/l glucose, 2 mM L-9

glutamine, 10 % heat-inactivated FBS (Gibco)) with 5 % CO2 at 37 ºC. We used plasmid pcDNA3 (Invitrogen) to 10 express CCs and its fusions with split enzymes. Renilla luciferase (phRL-TK, Promega) was used as a transfection 11 control in all experiments in mammalian cells. 12 13

Immunoblotting. 14 HEK293T cells were seeded in 6-well plates (Techno Plastic Products) at 2.5 × 10 4 cells per well (1 ml). HCl with a pH of 7.5, 150 mM NaCl, and 10 % glycerol) were filtered using Durapore 0.1 μm centrifuge filters 46

(Merck Millipore, Ireland) before being injected onto the Biosep SEC-S2000 column (Phenomenex, CA), with the 47 exception of the peptide N2, which was injected in a Superdex 30 increase column (GE, IL). The mobile phases 48 used for the separations were 50 mM Tris-HCl, pH 7.5, 150 mM NaCl and 50 mM Tris-HCl, pH 7.5, 500 mM 49

NaCl for the Biosep SEC-S2000 column and the Superdex 30 increase column, respectively. The injection volume 50 was 50 μl, and the flow rate was 0.5 ml/min. Data analysis was carried out with Astra 7.0 software (Wyatt, CA) 51 utilizing the theoretical extinction coefficient calculated by ProtParam. 52 53

1

MicroCal VP-ITC (Microcal, Northampton, MA) was used to measure the equilibrium dissociation constant (KD) 2 of orthogonal CC pairs. A 280-µL syringe was filled with the peptide solution (10 µM in 50 mM Tris-HCl with a 3 pH of 7.5, 150 mM NaCl, and 1 mM TCEP). The samples were injected into an ITC cell (volume 1.3862 ml) filled 4 with a complimentary peptide (1 µM) in a matching buffer through a 27-step process. The injection volumes were 5 10 µl and were spaced 600 s apart, with the exception of the first priming injection of 2 µl with 3200 s spacing. 6

Before each ITC experiment, the sample and buffer solutions were degassed in an ultrasonic water bath. The 7

integration of heat pulses and model-fitting were performed with Origin 7.0 using a single-site binding model. 8 9

Modeling CC interactions and structures. 10

To evaluate the orthogonality of the designed CC sets, the interactions between all possible peptide pairs were 11 predicted by the scoring function introduced by Potapov et al. 31 . For each peptide pair, the interaction score was 12 calculated for three different relative alignments of the peptide chain. The amino acid sequences were either 13 completely aligned or staggered by one heptad in the reverse or forward direction. The lowest score was taken as 14 the interaction score of the peptide pair. 15

The Potapov scoring function ignores the contribution of amino acids at the b, c, and f positions and, therefore, 16 cannot be used to predict the impact of mutations in these positions on CC stability. where is the melting temperature predicted by bCIPA 47 ; CP is the peptide pair charge product; and Hel 21 is the number of residues in a peptide pair with helicity above 0.05, as predicted by Agadir 32 . 22

The model structures of designed CCs were built using the ISAMBARD modeling package 48 . Structural 23 parameters (superhelix radius, pitch, and Crick angle) were optimized using a genetic algorithm. The following 24 space of parameters was searched: superhelix radius 5 ± 1 Å, pitch 200 ± 60 Å, Crick angle at a position 26 ± 27º. 25

The relative shift along the superhelical axis was assumed to be zero. The internal energy was evaluated using the 26 BUDE force field 49 . 27 28

Split luciferase and split transcription factor reconstitution.

HEK293T cells were seeded in 96-well plates (Corning) at 2.5 × 10 4 cells per well (0.1 ml). The next day, cells 30

were transiently transfected with plasmids expressing CCs fused to split luciferase or plasmids expressing CC-31 split transcription factor (sequences are shown in Table S4 ), and phRL-TK (Promega) constitutively expressing 32

Renilla luciferase (5 ng per well, for normalization of transfection efficiency) using the PEI transfection reagent. 33

The total amount of DNA for each transfection was kept constant by adding appropriate amounts of the control 34 plasmid pcDNA3 (Invitrogen Syncytia is formed by the fusion of cells expressing SARS-CoV-2 S protein on the cell surface with 47 neighboring cells expressing ACE2 receptor leading to the formation of multi-nucleate enlarged cells. 48 For fusion assay, the spike expressing (donor) cells and ACE2 receptor-expressing (acceptor) cells were 49 mixed in 1:1 ratio. HEK293T cells were seeded in 6-well plates (Corning) at 5 × 10 4 cells per well (2 ml). The 50

next day, cells were transiently transfected with plasmids. The acceptor group was transfected with plasmids 51 encoding cLuc:CC, ACE-2 receptor, and TMPRSS-2. The donor group was transfected with nLuc:CC and SARS-52 CoV-2 S protein using the PEI transfection reagent. The total amount of DNA for each transfection was kept 1 constant by adding appropriate amounts of the control plasmid pcDNA3 (Invitrogen). 24h after transfection, cells 2 were detached from 6 well plates using a solution of PBS, EDTA (2,5mM). Wells were additionally washed with 3 DMEM, 10% FBS. Cells were then centrifugated (5min, 1500rpm), resuspended in 1ml DMEM, 10%FBS, and 4

counted using the automated cell counter, Countess. Both groups of resuspended cells were diluted to a final 5 concentration of 1×10 5 cells/ml, mixed in ratio 1:1, and seeded to 96-well plates. TMPRSS2 inhibitor Camostat 6 mesylate (1-5µM) or RBD (10-100µg/ml) were added to diluted cells expressing ACE2 receptor before mixing 7 them with cells expressing S protein. 8

Three hours after mixing both cell groups, luciferase activity was measured as previously described 3 , 10 µl of 9 0.5mM luciferase substrate luciferin and 2.5mM ATP was added per well before luciferase activity was measured 10

with Orion II microplate reader, Berthold Technologies. 11 12

Viral fusion protein-mediated syncytium formation-Flow cytometry. 13

For flow cytometer analysis, the cells were seeded in a 12-well plate (Costar) (1 ml, 2.5 × 10 5 cells/ml). At 70% 14 confluency, the donor HEK293T cells were transfected with plasmids expressing S protein, fluorescent protein, 15

iRFP NLS , and syncytia reporter split fluorescent protein, GFP1-10 (GFP1-10:N7, GFP1-10:P7 or GFP1-10). Acceptor 16 cells were transfected with plasmids expressing a fluorescent protein, BFP NLS , ACE2 receptor, and syncytia 17

reporter split protein, GFP11 (N8:GFP11, 3×(N8:GFP11), P8:GFP11 )using PEI transfection reagent. Plasmid 18 pcDNA3 was used to adjust the amount of plasmid DNA. Eighteen hours post-transfection, the spike expressing 19 cells were detached from the surface using PBS, EDTA (2.5mM). Wells were additionally washed with DMEM, 20 10% FBS. Donor cells were then added to the ACE2 acceptor cells in 1:1 ratio. Syncytia were analyzed 3 h later. 21 For inhibition of syncytium formation, soluble RBD (50-100µg/ml) or Camostat (50-100µM) was added to the 22 cells at the same time as a mixture of cells was formed. 23

Syncytia formation was analyzed using the spectral flow cytometer Aurora with the blue, violet, and red lasers 24 (Cytek). The cells gated positive for iRFP and BFP were examined for fluorescence intensity of reconstituted split 25 GFP. The fluorescence intensity is presented as the median value. Data were analyzed using FlowJo (TreeStar, 26 Ashland, OR, USA) and SpectroFlo (Cytek) software.

Viral fusion protein-mediated syncytium formation-confocal microscopy.

For confocal microscopy, acceptor cells were seeded in an 8-well tissue culture chamber (µ-Slide 8 well, Ibidi 30

Integrated BioDiagnostics, Martinsried München, Germany) (0.2 ml, 2.5 × 10 5 cells/ml), and acceptor cells were 31 seeded in 24-well plates. At 70% confluency, the donor HEK293T cells were transfected with plasmids expressing 32 S protein, syncytia reporter split fluorescent protein GFP1-10 (GFP1-10:N7, GFP1-10:P7 or GFP1-10) and fluorescent 33 protein mCherry NLS , and the acceptor cells were transfected with plasmids expressing the ACE2 receptor, syncytia 34

reporter split protein GFP11 (N8:GFP11, 3×(N8:GFP11), P8:GFP11 ), and fluorescent protein BFP using PEI 35 transfection reagent. Plasmid pcDNA3 was used to adjust the amount of plasmid DNA. Eighteen hours post-36 transfection, the spike expressing cells were detached from the surface using EDTA (2,5mM) and added to the 37 ACE2 acceptor cells in 1:1 ratio. Syncytia were analyzed 3 h later. 38

Microscopic images were acquired using the Leica TCS SP5 inverted laser-scanning microscope on a Leica DMI 39 6000 CS module equipped with an HCX Plane-Apochromat lambda blue 63 × oil-immersion objectives with NA 40 1.4 (Leica Microsystems, Wetzlar, Germany). A 488-nm laser line of a 100-mW argon laser with 10% laser power 41 was used for split GFP excitation, and the emitted light was detected between 500 and 540 nm. A 1-mW 543-nm 42

HeNe laser was used for mCherry:N6:NLS transfection control excitation and emitted light was detected between 43 580 and 620 nm. A 50-mW 405-nm diode laser was used for BFP transfection control, and emitted light was 44 detected between 420 and 460 nm. The images were processed with LAS AF software (Leica Microsystems) and 45

ImageJ software (National Institute of Mental Health, Bethesda, Maryland, USA). 46 47

Dual-luciferase assays. 48

At indicated time points, the cells were lysed in Passive Lysis 1× Buffer (Promega) and analyzed with a dual-49 luciferase reporter assay to determine the firefly luciferase and the Renilla luciferase activities (Orion II microplate  50 reader, Berthold Technologies). Relative luciferase units (RLU) were calculated by normalizing the firefly 51 luciferase value to the constitutive Renilla luciferase value in each sample. Normalized RLU values (nRLU) were 52 calculated by normalizing the RLU values of each sample to the value of the indicated sample within the same 1 experiment. 2 3

Software and statistics. 4

Graphs were prepared with Origin 8.1 (http://www.originlab.com/), and GraphPad Prism 5 5 (http://www.graphpad.com/) was used for statistical purposes. An unpaired two-tailed t-test (equal variance was

Coiled coil protein origami: from modular design 3 principles towards biotechnological applications

A tunable orthogonal coiled-coil interaction 5 toolbox for engineering mammalian cells

Design of fast proteolysis-based signaling and logic circuits in mammalian cells

A tunable coiled-coil interaction toolbox for engineering 9 interactions and regulation of mammalian cells

Guiding Biomolecular 11 Interactions in Cells Using de Novo Protein-Protein Interfaces

Programmable T7-based synthetic transcription factors. Nucleic Acids 13 Res

Downsizing Proto-oncogene cFos to Short Helix-Constrained Peptides That Bind Jun

De Novo Designed Protein-17 Interaction Modules for In-Cell Applications

Design of a single-chain polypeptide tetrahedron assembled from coiled-coil segments

Design of coiled-coil protein-origami cages that self-assemble in vitro and in vivo

Progress Toward the Clinical Translation of Bioinspired 23 Peptide and Protein Assemblies

Design of fast proteolysis-based signaling and logic circuits in mammalian cells

A synthetic coiled-coil interactome provides heterospecific 29 modules for molecular engineering

Designing Heterodimeric Two-stranded α-Helical Coiled-coils

Modulation of Coiled-Coil Dimer 33 Stability through Surface Residues while Preserving Pairing Specificity

De novo design of orthogonal peptide pairs forming parallel coiled-coil 36 heterodimers

Designed heterodimerizing leucine zippers with a ranger of pIs and stabilities up to 10-15 M. 38

Vitro and in Vivo Specifications of Heterospecific Coiled-Coil Interaction Domains

Beat COVID-19 through innovation. Science (80-. )

The spike protein of SARS-CoV -A target for vaccine and therapeutic development

Structural Basis for Potent Neutralization of Betacoronaviruses by Single-Domain 46

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a 48 Clinically Proven Protease Inhibitor

Characterization of spike glycoprotein of SARS-CoV-2 on virus entry and its immune cross-50 reactivity with SARS-CoV

Matthias Tenbusch, Frank Neipel, A. S. H. 52 SARS-CoV-2 and SARS-CoV spike-mediated cell-cell fusion differ in the requirements for receptor 53 expression and proteolytic activation and are not inhibited by Bromhexine

Furin 4 cleavage of SARS-CoV-2 Spike promotes but is not essential for infection and cell-cell fusion

A bimolecular 7 multicellular complementation system for the detection of syncytium formation: A new methodology for 8 the identification of Nipah virus entry inhibitors

N@a and N@d: Oligomer and Partner Specification by Asparagine in Coiled-Coil 10 Interfaces

The contribution of buried polar groups to the 12 conformational stability of the GCN4 coiled coil

Conformational Dynamics of 14 Asparagine at Coiled-Coil Interfaces

Data-driven prediction and design of bZIP coiled-coil 16 interactions

Elucidating the folding problem of alpha-helices: local motifs, 18 long-range electrostatics, ionic-strength dependence and prediction of NMR parameters

Model-based thermodynamic analysis of reversible unfolding 21 processes

CRISPR-mediated modular RNA-guided regulation of transcription in eukaryotes

Biochemical Analysis of the Kruppel-associated Box (KRAB) Transcriptional Repression 27 Domain

A bistable genetic switch based on designable DNA-binding domains

Designable DNA-binding domains enable construction of logic circuits in mammalian 31 cells

The New State of the Art: Cas9 for Gene Activation and Repression

Coronaviruses post-SARS: update on replication and pathogenesis

Camostat mesylate inhibits 39 SARS-CoV-2 activation by TMPRSS2-related proteases and its metabolite GBPA exerts antiviral 40 activity

Versatile protein tagging in cells with split fluorescent protein

Syncytia formation by SARS-CoV-2-infected cells

Engineered ACE2 receptor traps potently neutralize SARS-CoV-2

Truncated human angiotensin converting enzyme 2; a potential 48 inhibitor of SARS-CoV-2 spike glycoprotein and potent COVID-19 therapeutic agent

Protein-Protein Interaction Detection: Methods 51 and Analysis

Selectional and mutational scope of 53 peptides sequestering the Jun-Fos coiled-coil domain

CCBuilder 2.0: Powerful and accessible coiled-coil modeling. Protein 2 Sci

High performance in silico virtual drug 4 screening on many-core processors

assessed with the F-test, assuming normal data distribution) was used for statistical comparison of the data. If not 7 otherwise indicated, each experiment was independently repeated at least three times. Each experiment was 8 performed in at least three biological parallels. 9 10Data availability.

The authors declare that the data supporting the findings of this study are available within the paper and its 12supplementary information files. The raw data are available from the corresponding author upon reasonable 13 request. 14 15Abbreviations.