The Glycan-Specificity of the Pineapple Lectin AcmJRL and its Carbohydrate-Dependent Binding of the SARS-CoV-2 Spike Protein

The current SARS-CoV-2 pandemic has become one of the most challenging global health threats, with over 530 million reported infections by May 2022. In addition to vaccines, research and development have also been directed towards novel drugs. Since the highly glycosylated spike protein of SARS-CoV-2 is essential for infection, it constitutes a prime target for antiviral agents. The pineapple-derived jacalin-related lectin (AcmJRL) is present in the medication bromelain in significant quantities and has previously been described to bind mannosides. Here, we elucidated its ligand specificity by glycan array analysis, quantified the interaction with carbohydrates and validated high-mannose glycans as preferred ligands. Because the SARS-CoV-2 spike protein was previously reported to carry a high proportion of high-mannose N-glycans, we tested the binding of AcmJRL to recombinantly produced spike protein. We could demonstrate that AcmJRL binds the spike protein with a low micromolar KD in a carbohydrate-dependent fashion, suggesting its use as a potential SARS-CoV-2 neutralising agent.


43
Since the end of 2019, the world is facing the severe acute respiratory syndrome corona-44 virus 2 (SARS-CoV-2) pandemic. SARS-CoV-2 is a novel coronavirus that rapidly spread 45 all over the world and infected over 530 million people so far (May 2022). It can infect the 46 respiratory tract and potentially results in the coronavirus-associated disease  Especially for older or immunocompromised patients, COVID-19 is likely to be lethal. So 48 far, more than 6.3 million deaths were reported in association with SARS-CoV-2. 49 A variety of novel and very potent vaccines entered the market at the end of 2020. 50 Vaccination is an indispensable approach to protect society from a SARS-CoV-2 infection. 51 However, a rising fraction of vaccinated individuals suffers from severe syndromes after 52 infection due to a continuous adaptation of the virus. Furthermore, a very small fraction of 53 people can not be vaccinated due to medical preconditions (1). Currently, first antiviral 54 drugs like Remdesivir and Ritonavir/Nirmatrelvir (Paxlovid) are getting established in SARS-55 CoV-2 therapy and more are under investigation since we need novel pharmaceutical 56 therapies to cure or prevent infections with SARS-CoV-2. 57 Coronavirus spike proteins (S-proteins) are essential for the infection process, they are 58 trimerizing fusion proteins that consist of the two subunits S1 and S2. It has been shown 59 that S-proteins have a complex and extensive glycosylation pattern (2), and coronavirus 60 spike proteins typically contain between 23-38 N-glycosylation sites (3) per protomer with 61 a significant population of oligomannose-type glycans (30%) (2). The elucidation of the 62 glycosylation pattern (3) of the SARS-CoV-2 S-protein has been essential for the 63 development of an effective vaccine (4). A site-specific glycan analysis by mass 64 spectrometry has revealed that the 22 glycosylation sites on the S-protein monomer are 65 occupied with a mixture of oligomannose-type, hybrid-type and complex-type glycans. The 66 content of exclusively oligomannose-type glycosylation sites was determined to be 28%, 67 which is above the level of typical host glycoproteins (3). However, it is less than for other 68 viral glycoproteins like HIV-1 Env, where the amount of oligomannose-type glycans was 69 found to be around 60% (3). N-glycosylation is not only vital for protein folding during 70 protein expression, but these glycans also shield antigenic epitopes of S-protein and allow 71 the virus to evade the host's immune system (5). Further, glycans are important ligands for 72 the first interaction with the host via its cell surface attachment receptors (5). For viral entry 73 in airway epithelial cells, the receptor-binding domain (RBD) of SARS-CoV-2 S-protein 74 binds angiotensin-converting-enzyme-2 (ACE2) with high affinity (6). 75 Given the extensive glycosylation of the S-proteins of coronaviruses, it has been 76 hypothezised that targeting surface glycans of S-proteins could lead to a decreased 77 virulence of SARS-CoV-2. In fact, a study revolving around 33 different plant carbohydrate-78 binding proteins, i.e. lectins, observed antiviral activity for 15 plant lectins against SARS-79 CoV and feline infectious peritonitis virus (FIPV) (7). Interestingly, the most prominent 80 antiviral properties were found for mannose-specific lectins, which might be related to 81 oligomannose-type glycans being essential for S-protein function. Further, Hoffmann et al. 82 showed the ability of mammalian lectins (e.g. CLEC4G, CD209) to block SARS-CoV-2 83 infection in vitro (8). 84 Drug repurposing is a particularly interesting approach due to the acute nature of the 85 current pandemic and led to the use of Remdesivir. Bromelain, a pineapple (Ananas 86 comosus) stem extract, is an approved drug that shows anti-edematous, anti-inflammatory 87 and fibrinolytic properties and is thus used to cure trauma-induced swelling (9), (10). 88 Proteases, peptidic protease inhibitors and the mannose-binding lectin AcmJRL (also 89 called AnLec) are the three main protein components of bromelain (11). It is likely that 90 proteases are responsible for the anti-inflammatory properties of bromelain. Additionally, 91 protease inhibitors prevent unspecific proteolysis as a safety mechanism that is slowly 92 removed during the intake of bromelain. A putative mode of action of AcmJRL remains to 93 be uncovered. 94 AcmJRL was recently characterised by Azarkan et al. (12) and its surprisingly high content 95 in bromelain was determined by Gross et al. (11). AcmJRL belongs to the family of Jacalin-96 related lectins (JRL) (13). One of the first representatives of this family is jacalin, the lectin 97 isolated from jackfruit (Artocarpus integrifolia). The JRL family can be divided in two main 98 classes according to their ligand specificity (13). Galactose-specific JRL (gJRL) are found 99 almost exclusively in the Moraceae plant family, most typically in the seed. Structurally, 100 those JRLs are tetramers of four identical protomers, each containing one carbohydrate 101 binding site. The complex biosynthesis of mature gJRLs includes co-and post-translational 102 modifications from one preprotein including N-glycosylation. On the other hand, mannose-103 specific jacalin-related lectins (mJRL) are found in various plants. The structure of 104 mannose-specific JRLs is less complex as they usually consist of two, four or eight 105 unprocessed peptides. Due to the absence of a signal peptide, they are considered as 106 cytoplasmic proteins. 107 Isothermal titration calorimetry experiments with AcmJRL revealed rather low binding 108 affinity towards D-mannose (Ka = 178 M -1 ), D-glucose (Ka = 83 M -1 ) and GlcNAc (Ka = 88 M -109 1 ) (12). On the other hand, oligomannose structures like mannotriose (Man-α-1,6(Man-α-110 1,3)Man) and mannopentaose (Man-α-1,6(Man-α-1,3)Man-α-1,6(Man-α-1,3)Man) showed 111 significantly higher binding affinities of Ka = 734 M -1 to 1694 M -1 . 112 Like other mannose-specific JRLs, AcmJRL adopts a characteristic β-prism fold and two 113 monomers align side-by-side forming a dimer. Although a tetrameric form of AcmJRL could 114 also be assigned from the monomers in the asymmetric unit, this is likely a crystallization 115 artefact (see below). In co-crystal structures with D-mannose and methyl α-D-116 mannopyranoside, two carbohydrates are bound by one monomer in a conserved binding 117 pose. Overall, the interactions are comparable to the binding of D-mannose by BanLec, a 118 closely related mannose-specific JRL from banana (Musa acuminata). BanLec is reported 119 as a potent viral entry inhibitor of HIV-1, HCV and influenza virus (14,15). However, the 120 mitogenic activity of native BanLec limits its therapeutic use. Interestingly, the structure of 121 AcmJRL shares similarities with a genetically engineered BanLec (15)  Here, the ligand specificity of AcmJRL was further characterised with two glycan arrays 125 and the interaction was quantified in a competitive binding assay. In a next step, we 126 demonstrated the ability of AcmJRL to bind to the SARS-CoV-2 spike protein as well as its 127 isolated receptor-binding domain (RBD) with low micromolar affinity in a carbohydrate-128 dependent manner. Further, we showed that the binding of the spike RBD to its receptor 129 ACE2 can be inhibited by AcmJRL. Consequently, it is possible that the mannose-binding 130 lectin AcmJRL can neutralise the SARS-CoV-2 virus through binding to its spike protein. Isolation and characterisation of AcmJRL from bromelain 135 The mannophilic lectin AcmJRL was isolated by affinity chromatography from pineapple 136 stem extract (bromelain) on a mannosylated stationary phase according to the procedure 137 reported by Azarkan et al. (12). Prior to purification, the soluble protein fraction of bromelain 138 was obtained by aqueous extraction in presence of S-methyl methanethiosulfonate to 139 block bromelain's high proteolytic activity. Using mannosylated sepharose beads (16) 0.9 140 -1.6 mg AcmJRL were obtained per gram bromelain after elution with mannose. 141 The identity of the isolated protein was confirmed by SDS-PAGE ( Figure 1A) and mass 142 spectrometry (average molecular mass = 15346 Da, Figure 1B). The main peak (m/z = 143 15388) can be assigned to an acetonitrile adduct [M+H+MeCN] + of the AcmJRL monomer. 144 As reported by Gross et al. (11), two additional mass peaks were observed in a ratio of 100 145 : 65 : 17, separated by a mass shift of 162 Da. During the industrial production of bromelain, 146 the pineapple stem extract is loaded on maltodextrin, a hydrolysis product of starch, to 147 simplify its handling. Thus, it contains carbohydrates like glucose and glucooligo-148 saccharides that can react with primary amines of proteins to form Schiff-bases, which is 149 presumably followed by an irreversible Amadori rearrangement ( Figure S1) towards a stable 150 α-amino ketone corresponding to advanced glycation end-products (17). This glycation 151 results in a mass increase of 162 Da that was observed in the MS-spectrum ( Figure 1B).

152
The presence of two signals of +162 Da and +324 Da suggests that this reaction occurred 153 twice on the protein or one disaccharide of maltose reacted. However, it is not clear if it is 154 a statistical mixture or if two specific lysines were affected by this reaction (5 lysines are 155 present in AcmJRL). It was not specified whether the AcmJRL isolated and crystallised by 156 Azarkan et al. was also glycated. Inspection of the electron density map of crystallised 157 AcmJRL (PDB: 6FLY (12)) did not show evidence for unassigned electron density which 158 could also result from a high flexibility at the protein surface or a statistical distribution of 159 the glycation.

169
Previous studies showed dimerisation of AcmJRL in solution, determined by size exclusion 170 chromatography and equilibrium unfolding experiments (12). However, Azarkan et al. 171 showed that AcmJRL crystallises as a tetramer (see introduction). We therefore used 172 dynamic light scattering (DLS) to determine the hydrodynamic diameter of AcmJRL in 173 buffered solution ( Figure 1C). The measured hydrodynamic diameter of 57 ± 9.6 Å 174 corresponds to the radius of the dimer ( Figure S2), rather than to monomers or tetramers. 175 Additionally, our differential scanning fluorimetry studies suggested two unfolding events 176 at T1 = 58 -60°C and T2 = 73 -74°C ( Figure S2) which could reflect dissociation of the 177 dimer followed by protein denaturation. 178

216
The distance from the binding site of one monomer to the other monomer in the crystal 217 structure is approx. 46-50 Å. The distance between the anomeric carbons of two 218 mannosides within one binding site is approx. 14 Å (from C1 to C1), which is similar to the 219 distance between two mannoses in mannopentaose suggesting a possible chelation 220 binding mode by this ligand. On the other hand, mannopentaose could also preorganise 221 two α-mannosides in a way that allows the rapid rebinding with the two binding sites within 222 one monomer. 223 224 225 226 Semiotik glycan ID and short name can be found in tables S1 and S2. Sp10 310 Sp12 213 Sp9 207 Sp12 214 Monovalent α-glucosides showed very low, but significant binding (e.g. CFG-GLYCAN ID 232 195, Glc-α-1,4-Glc-β, RFU = 33 ± 5). Unfortunately, no multivalent glucosides are available 233 on this array to understand the influence of multivalency for these epitopes. 234 In addition to the CFG glycan array, we analysed AcmJRL on the Semiotik glycan array (20), 235 which also features mammalian glycans and additionally a large variety of other glycans, 236 mainly from bacterial species ( Figure 2B). Although mannopentaose (Semiotik glycan ID 237 (SGID) 454) and poly-Man-α-1,6 (mannan, SGID 3002) could be confirmed as a ligand for 238 AcmJRL, the smaller mannotriose (SGID 258) showed no binding, arguably due to a shorter 239 linker length (Sp4) preventing accessibility for the protein. In contrast to the CFG glycan 240 array, a multivalent α-glucoside is present on the Semiotik chip (SGID 2208) and was well 241 recognized by the lectin, underlining the affinity of AcmJRL towards α-glucosides. 242 Interestingly, two unrelated bacterial O-antigens were also recognised: showed a nonlinear dose-response, which requires orthogonal binding assays for 246 verification. 247 248

286
Expression of SARS-CoV-2 Spike Protein 287 Given the potential antiviral properties of plant lectins, our characterisation of AcmJRL 288 made us speculate about the binding of AcmJRL to the heavily glycosylated SARS-CoV-2 289 spike protein. 290 The spike-protein of SARS-CoV-2 was recombinantly produced in HEK293 cells using a 291 pCAGGS-based eukaryotic vector system yielding 1.42 mg glycoprotein from one batch of 292 560 mL. Identity and purity of the protein were determined by gel electrophoresis ( Figure  293 5A) and mass spectroscopy ( Figure 5B). Both experiments coherently show a molecular 294 weight around 145 kDa. Interestingly, mass spectroscopy revealed four major masses after 295 maximum entropy (MaxEnt) deconvolution of the centroided mass spectrum, each 296 separated by approximately 2 kDa ( Figure 5B). As described above, S-protein usually 297 exhibits highly complex glycosylation, which leads to a non-homogeneous sample. 298 Therefore, the heterogeneous mass distribution determined by ESI-MS presumably 299 resulted from the presence of different glycoforms. 300 301 302   Peptide-N-glycanase F (PNGase F) is an amidase that specifically hydrolyses amide bonds 314 between the reducing end GlcNAc and asparagine residues of N-glycans. Consequently, 315 PNGase F was used to verify the glycosylation of the recombinantly produced N-316 glycosylated S-protein. S-protein was incubated in the presence of PNGase F and 317 compared with the untreated protein sample ( Figure 5). In fact, the treatment resulted in a 318 faster migration on the SDS-gel indicating a molecular weight reduced by several kDa 319 ( Figure 5A). Coherently, the MaxEnt deconvoluted mass spectrum of the treated S-protein 320 shows one major peak around 138 kDa, together with two small satellite peaks. Notably, 321 all mass peaks of the PNGase F-treated species were much sharper and reached higher 322 signal intensities compared to the native protein. The lower intensity of the deconvoluted 323 mass peaks of the untreated protein in comparison to the higher intensity of the 324 deconvoluted mass peaks of the PNGase F-treated protein indicated that a smaller 325 diversity of glycoprotein species is existent in the deglycosylated sample while in the 326 glycosylated native sample, the presence of a complex mixture of different glycoforms is 327 responsible for the lower peak intensities due to the lower abundance of each individual 328 species. The most intense peak in the deconvoluted mass spectra reveals a mass shift of 329 approximately 7 kDa, resulting in a main mass peak around 138 kDa. Therefore, the mass 330 spectrometric analysis of the enzyme treated sample qualitatively confirmed an extensive 331 glycosylation of the recombinantly produced glycoprotein. 332 333

Mannose-dependence of AcmJRL binding to SARS-CoV-2 spike-protein 334
After verification of the S-protein identity, a biophysical assay for determination of binding 335 kinetics and affinity of AcmJRL to the S-protein was established. Given the high molecular 336 weights of both interaction partners, Surface Plasmon Resonance (SPR) analysis was 337 chosen to determine binding affinity and kinetics of AcmJRL against immobilised S-protein.

338
Single cycle kinetics were performed by injecting AcmJRL at increasing concentrations 339 from 2.5 to 20 µM onto a sensor chip with immobilised S-protein and a dose-dependent 340 response was observed. Even at the highest concentration, saturation could not be 341 observed after 120 sec association, hinting at an extensive number of available binding 342 sites for AcmJRL. In addition, only an incomplete dissociation was recorded during the 343 dissociation phases. The association rate (kon) was 1057 ± 94 M -1 s -1 , the dissociation rate 344 (koff) was 3.42 ± 0.4 x 10 -3 s -1 which corresponds to a KD of 3.27 ± 0.5 µM. A similar KD of 345 11.1 ± 2.5 µM was determined after fitting the response after 115 sec contact time of the 346 association with the Langmuir isotherm, which is more error prone due to the fact that 347 saturation was not reached. 348 Interestingly, the KD increased gradually for each single experiment performed for the 349 technical replicates. (Dissociation constants calculated from Langmuir isotherm: 7.9 µM, 350 11.5 µM, 14.0 µM; Dissociation constants calculated from rate constants: 2.7 µM, 3.1 µM, 351 4.0 µM). This observation likely resulted from some AcmJRL remaining bound during the 352 regeneration cycles as the S-protein has a vast number of possible binding sites in its 353 numerous N-glycans for AcmJRL. Furthermore, during dissociation phases the RU 354 (response units) values did not fully decrease to the baseline response, indicative for an 355 incomplete dissociation of AcmJRL. 356 As AcmJRL interacts with the mannotriose epitope (Man-α-1,6(Man-α-1,3)Man, present in 357 CFG glycan ID 211, 213, 51 and 50) in the glycan array analysis, single cycle kinetics of 358 AcmJRL with the addition of 10 mM mannotriose in the sample buffer were also conducted 359 to analyse the glycan-dependence of the AcmJRL-spike binding. The loss of observable 360 binding to the spike protein in presence of the competitor demonstrates the mannose-361 dependent binding of AcmJRL ( Figure 6). 362 363 364 365

370
AcmJRL binds to SARS-CoV-2 spike RBD and to its native receptor ACE2 371 The RBD of the S-protein is essential for the SARS-CoV-2 infection process, as it mediates 372 binding to the human cell surface receptor ACE2, allowing the attachment of the virus. As 373 we observed the binding of AcmJRL to the S-protein, we set out to determine if the lectin 374 is able to block this essential mechanism for the infection through binding to the spike 375 protein. Complex glycosylation is present in many other human cell surface proteins, such 376 as the receptor ACE2. 377 Consequently, we also analysed the binding of AcmJRL to ACE2 as well as to spike RBD 378 produced for interaction studies. Single cycle kinetics on SPR were performed for AcmJRL 379 binding to both the S-protein RBD and the human ACE2 receptor (Figure 7). Twentyone response [RU] AcmJRL vs spike 455 M -1 s -1 ) were comparable to those for AcmJRL binding to S-protein ( Figure 6). However, 387 the dissociation of AcmJRL from spike RBD is about fourfold faster, displayed in the 388 determined koff = 19.4 ± 6.7 x 10 -3 M -1 s -1 , a consequence of the reduced extent of 389 glycosylation of the RBD.

397
Human ACE2 is also highly glycosylated, carrying more matured complex glycans 398 compared to the high-mannose enriched virus surface protein (27). From SPR analysis of 399 AcmJRL binding to immobilised recombinant ACE2 produced in HEK293 cells, we obtained 400 a KD value of 22.1 ± 5.3 µM calculated from the Langmuir isotherm, which is sevenfold 401 higher than the one obtained for binding to the S-protein. This observation could be a direct 402 result of the altered glycosylation pattern of this human receptor that is distinct from the 403 viral proteins. The association rate of AcmJRL to ACE2 (kon= 590 ± 212 M -1 s -1 ) was also 404 slower compared to the one for the S-protein. Further, the threefold higher dissociation rate 405  AcmJRL vs ACE2 (koff 12.2 ± 1.0 x 10 -3 M -1 s -1 ) of AcmJRL from the ACE2 receptor indicates a faster 406 dissociation of the complex. In contrast to the interaction with full length S-protein, 407 complete dissociation of AcmJRL from both the RBD and ACE2 complexes was observed. 408 The observed affinity of AcmJRL for both S-protein and ACE2 could therefore result in a 409 synergistic inhibitory effect on viral cell entry.

411
AcmJRL weakens the interaction of spike RBD with ACE2 412 The interaction between the spike RBD and its ACE2 receptor is characterised by 413 nanomolar affinity and is essential for the infection process. As AcmJRL binds the S-protein 414 RBD and ACE2 with low micromolar affinity, it is likely that the spike interaction with ACE2 415 could be inhibited by AcmJRL. To test this hypothesis, we first reproduced the affinity 416 reported for the S-protein RBD to the immobilised ACE2 receptor (Figure 8)

425
The influence of AcmJRL on the affinity of S-protein RBD binding to the ACE2 receptor was 426 studied by an addition of AcmJRL with a molar excess of lectin (factor 10 and 100, i.e. 1 427 µM and 10 µM Figure 8) to the RBD and preincubation prior to injection. Then, a single 428 cycle kinetics run with injections of 5 dilutions (6.25 nM RBD with 10 and 100 fold excess 429 AcmJRL -100 nM RBD with 10 and 100 fold excess AcmJRL) of the RBD-AcmJRL mixture 430 was performed with immobilised ACE2 receptor. Although residual binding of RBD to ACE2 431 receptor could still be observed (Figure 8), its apparent affinity was reduced by a factor of 432 two (KD = 21.1 ± 2.9 nM) after preincubation with a 10-fold excess of AcmJRL. An increase 433 AcmJRL AcmJRL to a 100-fold excess of preincubated AcmJRL led to another twofold decrease in KD value 434 to 37.1 ± 4.7 nM. This rather moderate inhibitory effect probably resulted from the dilution 435 of the preincubation mixture into the different injected concentrations, resulting in AcmJRL 436 concentrations below the determined Kds of AcmJRL for spike RBD and ACE2. The 437 addition of sufficiently high concentrations of AcmJRL to saturate RBD could not be used 438 to overcome this problem, since AcmJRL also directly binds to the glycans of the 439 immobilized ACE2 impacting on the recorded SPR response. 440 In a clinical scenario, saturating both the glycans of ACE2 and spike with AcmJRL could 441 be beneficial to weaken the virus-host interaction and provide the immune system with an 442 added advantage while battling the infection. 443

446
The current SARS-CoV-2 pandemic is a serious crisis that urgently asks for therapeutic 447 treatment options. The viral envelope of SARS-CoV-2 is densely covered by the highly 448 glycosylated spike protein, which is essential for viral cell entry via binding to the ACE2 449 receptor. 450 In this work, the pineapple-derived jacalin-related lectin AcmJRL was purified from the 451 active pharmaceutical ingredient bromelain and characterised by mass spectrometry, 452 differential scanning fluorimetry and dynamic light scattering. We further analysed the 453 lectin's ligand specificity by glycan array analysis using two complimentary arrays. The data 454 further supported the previously reported preference of AcmJRL for mannopentaose. A 455 solution phase binding assay was subsequently developed to quantify AcmJRL-456 carbohydrate interactions. 457 Then, the interaction of AcmJRL to recombinantly produced SARS-CoV-2 spike protein 458 was studied by surface plasmon resonance analysis. The low µM binding was 459 carbohydrate-dependent and could be inhibited by supplementation with mannotriose. 460 Finally, we could show that addition of AcmJRL reduced the tight binding affinity of the 461 spike RBD for the human ACE2 receptor. Thus, bromelain and specifically its component 462 AcmJRL could constitute a novel antiviral drug to neutralise SARS-CoV-2 post exposure. The authors are thankful to Ursapharm (Saarbrücken, Germany) for funding parts of this 467 work. We are indebted to Dr. Jamie Heimburg-Molinaro and Kelly Baker for the CFG glycan 468 array analysis and we acknowledge the participation of the Protein-Glycan Interaction 469 Resource of the CFG and the National Center for Functional Glycomics (

489
Preparation of D-mannosylated sepharose 490 D-Mannosylated sepharose was synthesised according to the protocol of Fornstedt and 491 Porath (16): Sepharose CL-6B beads (Sigma-Aldrich Chemie GmbH, Germany, 15 mL) were 492 suspended in Na2CO3-buffer (500 mM, pH 11, 15 mL). Divinylsulfone (1.5 mL) was added 493 and the suspension was stirred at r.t. for 70 min. Activated sepharose was extensively 494 washed with water and resuspended in 15 mL aqueous D-mannose solution (20% m/v, 500 495 mM Na2CO3, pH 10). The reaction was stirred over night at r.t., filtered and extensively 496 washed with water. Unreacted activated sepharose was quenched by addition of β-497 mercaptoethanol (300 µL) in buffer (15 mL, 500 mM NaHCO3, pH 8.5) for 120 min. After 498 filtration and washing of the mannosylated beads, they were filled into 5 mL plastic columns 499 for affinity chromatography. 500 501 Isolation of AcmJRL from bromelain 502 AcmJRL was isolated in analogy to the protocol of Azarkan et al. (12) . Bromelain powder (28 503 g) was suspended in an Erlenmeyer flask in buffer (400 mL, 100 mM NaOAc pH 5, 1 mM 504 EDTA, 20 mM methyl methanethiosulfonate) and stirred for 60 min at r.t.. After 505 centrifugation (30,000 rcf, 30 min, 4 °C), the supernatant was dialysed twice for 1 h against 506 4 L Tris-buffered saline (TBS: 150 mM NaCl, 50 mM Tris pH 7.4). The sample was loaded 507 on a D-mannosyl-sepharose column pre-equilibrated with the dialysis buffer. After 508 extensive washing, the lectin was eluted with 1 M D-mannose in buffer. Eluted fractions 509 containing AcmJRL were pooled and dialysed against TBS (5 x 3 h against 2 L). The yield 510 (31 mg) was determined by UV-absorption at 280 nm (MW = 15.34 kDa, ε = 19940 M -1 x 511 cm -1 ). 512 513 Intact protein mass determination 514 Intact protein mass measurements for AcmJRL were performed on a Dionex Ultimate 3000 515 RSLC system using an Aeris Widepore XB C8, 150 x 2.1 mm, 3.6 μm dp column 516 (Phenomenex, USA). Separation of a 2 μL sample was achieved by a linear gradient from 517 (A) H2O + 0.1% formic acid to (B) MeCN + 0.1% formic acid at a flow rate of 300 μL/min 518 and 45 °C. The gradient was initiated by a 1 min isocratic step at 2% B, followed by a linear 519 increase to 75% B in 10 min to end up with a 3 min step at 75% B before re-equilibration 520 with initial conditions. UV spectra were recorded on a DAD in the range from 200 to 600 521 nm. The LC flow was split to 37.5 μL/min before entering the maXis 4G hr -ToF mass 522 spectrometer (Bruker Daltonics, Bremen, Germany) using the standard Bruker ESI source. 523 In the source region, the temperature was set to 200 °C, the capillary voltage was 4000 V, 524 the dry-gas flow was 5.0 L/min and the nebuliser was set to 1.0 bar. Mass spectra were 525 acquired in positive ionisation mode ranging from 150 -2500 m/z at 2.0 Hz scan rate. 526 Protein masses were deconvoluted by using the Maximum Entropy algorithm (Spectrum 527 Square Associates, Inc. Labtech GmbH, Germany) and polarization was calculated. The data were analysed using 554 a four-parameter fit calculated with MARS Data Analysis Software (BMG Labtech GmbH, 555 Germany). Three independent measurements on three plates was performed.

557
Reporter ligand displacement assay 558 The assay was performed in analogy to the protocol from Joachim et al. were removed by filtration (Vivaspin, 10000 MWCO), then the protein was affinity-purified 582 as described above for unlabelled AcmJRL. The protein concentration and degree of 583 labelling (DOL) was calculated according to the manufacturers protocol (Thermo Scientific, 584 Rockford NHS-activated Cy3: AcmJRL was diluted in PBS pH 8.4 and concentrated (Vivaspin, 595 Sartorius Stedim Biotech GmbH, 10,000 MWCO) to yield a final protein concentration of 596 293 µM (4.5 mg in 1.5 mL). NHS-activated Cy3 (Lumiprobe, Germany, 75 µL of a freshly 597 prepared 29 mM solution in DMSO, 2.2 µmol, 9 eq.) was added and incubated for 5 h at 598 r.t.. Excess reagents were removed by filtration (Vivaspin, Sartorius Stedim Biotech GmbH, 599 10000 MWCO), then the protein was affinity-purified as described above for unlabelled 600 AcmJRL. The protein concentration and degree of labelling (DOL) was calculated as 601 described above. Glycan array analysis 607 FITC-labeled AcmJRL was tested by the National Center for Functional Glycomics (NCFG, 608 Boston, MA, USA) on the CFG glycan microarray version 5.5 containing 585 printed glycans 609 in replicates of 6. Standard procedures of NCFG (details see 610 https://ncfg.hms.harvard.edu/files/ncfg/files/protocol-direct_glycan_binding_assay-611 cfg_slides.docx) were run with 5 and 50 µg/mL protein based on the protocol by Blixt et al. 612 (19). Raw-data (Tables S5 and S6) will be shared online on the CFG website. 613 Cy3-labelled AcmJRL was tested in-house on a glycan microarray slide from Semiotik LLC 614 (Moscow, Russia) containing 610 printed glycans in replicates of 6. Standard procedures 615 were run at 20, 200 and 400 µg/mL based on the protocol by . 616 Fluorescence intensity was measured at 565 nm upon excitation at 520 nm on a Sapphire 617 Biomolecular imager (Azure Biosystems, Dublin, CA, USA) at 10 µm resolution. Scan data 618 was processed with ScanArray software (Perkin Elmer, Waltham, MA, USA), using 619 OSPS090418_full.360.80 um.gal (kindly provided by Semiotik) for dot-glycan assignment. 620 Raw data (dot mean fluorescence intensity) was processed by GraphPad Prism 9 621 (GraphPad Software, USA). Processed data in table format is provided in the appendix.

635
Cloning and Recombinant Expression of SARS-CoV-2 spike protein 636 A synthetic DNA fragment was purchased from Eurofins MWG. The nucleotide sequence 637 was codon optimised for mammalian cell expression (translational amino acid sequence 638 based on PDB code: 6VXX (29)). The nucleotide sequence coding for the extraviral domain 639 of SARS-CoV-2 spike (coding for aa 1-1213) was amplified via PCR with the restriction 640 sites 5`-BamHI/XhoI-3` (Fw Primer: ATATGGATCCATGTTCGTGTTCCTGGTTCTT; Rv 641 Primer: AATATGAGCAGTACATAAAATGGCCCCTCGAGATAT; purchased from Merck). As 642 vector system, the in house vector πα-SHP-H (provided by Dr. Jesko Köhnke, pCAGGS 643 based, NCBI accession number: LT727518) was chosen for the mammalian expression 644 system. The amplicon was digested with the respective restriction enzymes (ThermoFisher) 645