Glycan-Induced Protein Dynamics in Human Norovirus P Dimers Depend on Virus Strain and Deamidation Status

Noroviruses are the major cause of viral gastroenteritis and re-emerge worldwide every year, with GII.4 currently being the most frequent human genotype. The norovirus capsid protein VP1 is essential for host immune response. The P domain mediates cell attachment via histo blood-group antigens (HBGAs) in a strain-dependent manner but how these glycan-interactions actually relate to cell entry remains unclear. Here, hydrogen/deuterium exchange mass spectrometry (HDX-MS) is used to investigate glycan-induced protein dynamics in P dimers of different strains, which exhibit high structural similarity but different prevalence in humans. While the almost identical strains GII.4 Saga and GII.4 MI001 share glycan-induced dynamics, the dynamics differ in the emerging GII.17 Kawasaki 308 and rare GII.10 Vietnam 026 strain. The structural aspects of glycan binding to fully deamidated GII.4 P dimers have been investigated before. However, considering the high specificity and half-life of N373D under physiological conditions, large fractions of partially deamidated virions with potentially altered dynamics in their P domains are likely to occur. Therefore, we also examined glycan binding to partially deamidated GII.4 Saga and GII.4 MI001 P dimers. Such mixed species exhibit increased exposure to solvent in the P dimer upon glycan binding as opposed to pure wildtype. Furthermore, deamidated P dimers display increased flexibility and a monomeric subpopulation. Our results indicate that glycan binding induces strain-dependent structural dynamics, which are further altered by N373 deamidation, and hence hint at a complex role of deamidation in modulating glycan-mediated cell attachment in GII.4 strains.

formation of an iso-aspartate (iD) in GII.4 Saga P dimers that strongly attenuates glycan binding. 83 This deamidation appears to be site specific and occurs in GII.4 MI001 P dimers as well, whereas 84 it is absent in GII.10 Vietnam 026 and GII.17 Kawasaki 308 P dimers, which carry an Asp at the 85 equivalent position [19]. HDX-MS measurements confirmed the loss of binding of deamidated 86 P dimers to HBGA B trisaccharide and revealed increased flexibility in the P2 domain compared 87 to the wildtype P dimer. 88 HDX-MS measures the exchange of protein backbone hydrogens to deuterium in solution. As this 89 exchange strongly depends on solvent accessibility and hydrogen bonding patterns, the method 90 can provide information about regions involved in ligand binding as well as changes in protein 91 dynamics in solution [20]. This makes it a valuable technique for identification of glycan induced 92 structural dynamics in different strains as well as elucidation of altered protein dynamics in 93 deamidated P dimers. While P dimers across strains are structurally highly similar, their glycan 94 binding behavior and infectivity is highly variable, leading to the hypothesis that varying structural 95 dynamics is causing these different profiles. 96 Therefore, we set out to examine whether glycan binding or deamidation can induce distinct 97 structural dynamics changes in P dimers, thereby modulating infectivity. We specifically 98 investigated binding of HBGA B trisaccharide and L-fucose to P dimers of GII.4 Saga,GII.4 MI001,99 GII.17 Kawasaki 308, and GII.10 Vietnam 026. GII.4 MI001 infects humans and mice [21] and has 100 been chosen as comparison to the almost identical strain GII.4 Saga. GII.17 Kawasaki 308 is an 101 emerging strain, and the less abundant GII.10 Vietnam 026 is capable of binding four fucose 102 molecules per P dimer. The structural aspects of glycan binding to fully deamidated GII.4 P dimers 103 have been investigated before [19]. However, in vivo, large fractions of partially deamidated P 104 dimers with potentially altered dynamics are likely to occur. Therefore, we also examined glycan 105 binding to partially deamidated GII.4 Saga and GII.4 MI001 P dimers.
Our targets share a different amount of sequence identity with the GII.4 Saga P dimer, but are 107 highly similar on the structural level ( Figure 1), with largest differences in the loop regions of the 108 P2 domain. The results reveal identical glycan binding behavior in GII.4 Saga and GII.4 MI001 109 strains but distinct glycan induced dynamics in GII.17 Kawasaki 308 and GII.10 Vietnam 026. 110 Furthermore, all strains apart from GII.4 Saga form a second P domain species that is highly 111 protected from HDX. In partially deamidated GII.4 P dimers, fucose binding leads to different 112 structural dynamics than in pure wildtype or fully deamidated samples, hinting at a potential 026 (VP1 residues 224-538), and GII.17 Kawasaki 308 (VP1 residues 225-530) P domains (see 130 Figure S1 for VP1 sequence alignment), with GenBank accession numbers AB447457, 131 KC631814, AF504671, and LC037415, respectively, were synthesized and purified as described 132 Transformed cells were grown for 3 h at 37 °C. Overexpression was induced with 1 mM isopropyl-138 β-D-1-thiogalactopyranoside (IPTG) at an OD600 value of 1.5. Incubation was continued at 16 °C 139 for 48 h. Cells were lysed using a high-pressure homogenizer (Thermo). The lysate was clarified 140 by centrifugation, and the fusion protein was purified using a Ni-NTA resin (Qiagen). MBP and the 141 His-tag were cleaved from the P domain using HRV 3C protease (Novagen). Cleaved P domain 142 protein eluted from Ni-NTA resin and was further purified by size-exclusion chromatography using 143 a Superdex 16/600 200 pg column (GE Healthcare) in 20 mM sodium phosphate buffer (pH 7.3). 144 Protein purity and dimer concentration were monitored by SDS-PAGE and ultraviolet absorption. 145 Separation of fully, partially, and non-deamidated (pure N373 wildtype) GII.4 P dimer species was 146 achieved by cation exchange chromatography using a 6 ml Resource S column (GE Healthcare) 147 at 6 °C. After separation protein samples were prepared in 20 mM sodium acetate buffer (pH 4.9) 148 to prevent further spontaneous deamidation and eluted using a linear salt gradient. 149 Wildtype P dimer samples were stored at 5 °C in the following buffers until analysis: GII.10 150  source. Gold-coated electrospray capillaries were produced in house for direct sample infusion. 173

Glycan structures
Capillary and sample cone voltages were 1.20 kV to 1.35 kV and 150 to 240 V, respectively. The 174 pusher was set to 100-150 µs. Pressures were 7 mbar in the source region and 6.2 x 10 -2 to 6.5 x 175 10 -2 mbar argon in the hexapole region. A spectrum of a 25 mg/ml cesium iodide solution from the 176 same day was applied for calibration of raw data using the MassLynx software (Waters, UK). 177 OriginPro 2016 (Origin Lab Corporation) software was used for peak integration and calculation 178 of oligomer fractions. 179 180 HDX-MS 181 P dimers (30-50 pmol) were mixed with glycans at tenfold of the final concentration (final: 10 mM 182 HBGA B trisaccharide, 100 mM fucose) and directly diluted 1:9 in 99% deuterated 20 mM Tris 183 buffer (pH 7, 150 mM NaCl, 25°C) to start the exchange reaction. After various time points the 184 exchange reaction was quenched by 1:1 addition of ice-cold quench buffer (300 mM phosphate 185 buffer, pH 2.3, 6 M urea), which decreased the pH to 2.3, and frozen in liquid nitrogen. As a fully 186 deuterated (FD) control, P dimers were diluted 1:9 in 99% deuterated 20 mM Tris buffer with 187 150 mM NaCl and 6 M urea at pH 7, labelled for 24-72 h at room temperature and quenched as 188 was only considered to have a significant HDX difference if the peptide passed the T-test and ΔD 230 exceeded 2x the pooled average standard deviation [23,24] of the dataset either for several time 231 points or for the same time point in overlapping peptides. For some peptides deuteration of FD 232 controls was lower than deuteration of the 8 h labeling time point. Therefore, datasets were not 233 normalized to the absolute FD deuterium uptake and only relative differences between states are 234 presented. For comparison of the unbound wildtype and deamidated MI001 P dimer, the ratio of 235 the FD controls from both measurements was used for normalization. Additionally, a higher cut-236 off of ΔD > 0.42 (99% percentile calculated according to [25]) was used to account for possible 237 day-to-day variation in the experimental conditions. Regions with significant deuterium uptake 238 differences were mapped to existing P dimer crystal structures or the homology model (GII.4 239

MI001). 240
Deuterated spectra of peptides in certain protein regions showed bimodal peak distributions that 241 led to lower deuteration values in centroid analysis. To validate the deuteration differences 242 observed in centroid data analysis and to calculate relative intensities of both peak distributions, 243 bimodal spectra of peptides representative for certain regions were analyzed by binomial fitting in 244 HXExpress [26]. To compare relative intensities of both distributions in different states, an average 245 over all bimodal time points for both distributions in each state was calculated for several peptides. 246 Averaged relative intensities of the first peak distribution in different peptides are presented as bar 247 plots in Figure 3. The statistical significance of relative intensity differences of the first peak 248 distribution in different states were analyzed with a two-sided Student's T-test for each pair of 249 states (unbound vs. ligand-bound) in an individual experiment (p < 0.05). Peptide coverage maps, 250 indicating the effective peptide coverage in each HDX experiment, were plotted with MS Tools 251 [27] and can be found in the supplement ( Figures S12-16). as well as deuterium uptake plots for each dataset can be found in the supplement (Table S3   (pdb 6H9V) (iDiD). All pdb-structures were refined by adding missing atoms and residues using 290 the UCSF Chimera tool (version 1.14) [36]. GII.4 Saga P dimers were additionally simulated with 291 α-L-methyl-fucose (F) ligands to explore a potential influence of deamidation on protein dynamics.
Hence, the amount of systems is expanded to include wildtype GII.4 Saga P dimers (pdb 4X7C) 293 (NN) with one (NFN) and two (NFNF) fucose ligands, iDN P dimers with one fucose complexing 294 each individual chain (iDFN and iDNF), and two fucoses (iDFNF), and further include iDiD P dimers 295 with one (iDFiD) and two (iDFiDF) fucose ligands. systems were based on the sidechains' pKa at pH 7. Each system was minimized using the 304 steepest descent algorithm, followed by a 100 ps simulation with applied position-restrains. 305 Temperature and pressure were maintained at 300 K and 1 bar by the v-rescale thermostat and 306 the Parrinello-Rahman barostat, with coupling constants of 50 fs for both [42][43][44]. Neighbor lists 307 were updated every 10 steps. The particle mesh Ewald algorithm was used for Coulomb 308 interactions, with a real-space cut-off of 1.0 nm [45,46]. The systems were allowed to relax for 309 100 ns with a 5 fs time step, extracting one frame every 10 th ns as starting structures for the 310 production runs. Final simulations were performed for ten 100 ns production runs at a 5 fs time 311 step. As such, each of the 14 systems was simulated in 10 replicates from different starting 312 structures, resulting in an aggregated simulation time of 1 µs per system, making 14 µs in total for 313 all systems. 314 The root-mean-square deviation (RMSD) and fluctuation (RMSF), as well as the solvent 315 accessible surface area (Asas), were calculated to analyze the behavior of each system. The 316 RMSD was computed with the first frame of the individual trajectory as reference structure. The 317 trajectories of the ten replicas were combined to a single trajectory, of which the average structure was calculated with the Gromacs software package. This average structure was then taken as 319 reference for RMSF calculations, as this most accurately represents the standard deviation of the 320 individual atomic positions. To further support the RMSF calculations, we computed the Asas of the 321 protein backbone for the initial conformation (pdb-structure) and the final production simulations, 322 of which latter was combined to an average representation of the area over all ten production 323 replicas. with MS-Viewer using the respective search keys given in the supplement (Table S4). 332

335
To verify the deamidation status P dimer samples were subjected to peptic cleavage followed by 336 LC-MS for peptide identification (Table S2). No deamidated peptides were identified for GII.10 337 Vietnam and GII.17 Kawasaki after 1 year and 4 months (Vietnam) and 1 year (Kawasaki) of 338 storage at 5°C, respectively. For GII.4 MI001 P dimers stored for 1 year at pH 7.3 and 5°C, a 339 fraction of approximately 64 % was deamidated at N373. Based on this the ratio of purely native 340 (NN) to half deamidated (iDN) to fully deamidated (iDiD) P dimers is statistically predicted as 341 13:46:41 %. Furthermore, a minor fraction was deamidated at N239 and N448, respectively. For 342 GII.4 MI001 P dimers stored at pH 4.9 for 5 months at 5°C, no deamidation of N373 was observed. 343 Only a small fraction (< 10 %) of deamidated N448 was identified. GII.4 Saga P dimers stored for 344 more than 2 years at pH 7.3 at 5°C were approximately 88 % deamidated at N373 leading to a 345 ratio of NN:iDN:iDiD of 1.5:21:77.5 % (Table S1 and Figures S3-9). 346 Prior to HDX-MS analysis P dimers were subjected to native MS for quality control. Furthermore, 347 ion exchange separated wildtype (NN) and fully deamidated (iDiD) GII.4 Saga P dimers were 348 measured for comparison. GII.17 Kawasaki, GII.10 Vietnam, wildtype GII.4 MI001 and wildtype 349 GII.4 Saga P domains showed dimers with the expected molecular masses, apart from a small 350 fraction of unspecific tetramers formed during the ESI process ( Figure S10) experiments to P dimers of other human norovirus strains to analyze possible strain specific 361 differences in the structural response to glycan binding. Therefore, we incubated GII.4, GII.10 and 362 GII.17 P dimers with 10 mM HBGA B trisaccharide or 100 mM fucose at pH 7 and measured 363 differences between the unbound and ligand-bound state using HDX-MS. During inspection of the 364 glycan binding data, we observed that deuterated spectra of some peptides had a bimodal 365 character with one low intense, low deuterated and a second high intense, higher deuterated peak 366 distribution ( Figure 2A). These bimodal peak distributions can have many causes, e.g. two distinct 367 protein conformations, conformational rearrangements that lead to EX1 exchange kinetics, 368 insufficient ligand saturation or peptide carry over from the analytical or protease column [50,51]. 369 To rule out effects induced by carry over, an additional dataset with randomized sample order and 370 additional washing of the pepsin column between sample injections was measured for GII.4 MI001 371 wildtype (wt, N373) P dimers incubated with fucose, which still showed bimodality. Moreover, 372 bimodality is also observed in absence of any ligand, strongly indicating that undersaturated 373 binding sites are not the origin of bimodality. Furthermore, ligand concentrations were chosen to 374 shift perturbation titrations demonstrate the presence of two independent HBGA binding sites [19,49]. The presence of four fucose binding sites shown for GII.10 Vietnam P dimers in the presence 383 of high fucose concentrations appears to be an exception [12]. 384 Bimodality can be seen for P dimers of GII.4 MI001, GII.10 Vietnam and GII.17 Kawasaki in the 385 unbound proteins as well as in ligand bound forms. However, no bimodality is observed for GII.4 386 Saga P dimers [19]. For all three strains, bimodality is almost exclusively present in the P2 domain 387 and in the lower part of the P1 domain ( Figure 2B). Peptides that are affected by glycan binding 388 are often bimodal as well; therefore, it was necessary to manually analyze deuteration differences 389 in these regions again by binomial fitting of the individual peak distributions (an example analysis 390 can be seen in Figure S11). Residues 334-354, for example, are bimodal and also involved in highly similar for peptides within the same protein, which lead us to the assumption that the 396 P dimer adopts two distinct conformations, a compact and a more flexible one. The relative 397 intensity ratios of the peak distributions vary between experiments, but can still be compared within 398 a certain experiment ( Figure 2C). Depending on the strain and the experiment, the relative 399 intensity of the first peak distribution in the unbound P dimer varies between 7 and 17 %. For Kawasaki P dimers, significant protection in presence of HBGA B trisaccharide is only present in 441 this specific region. When incubated with fucose, additional protection of the canonical glycan 442 binding site (G443, Y444) and residues 269-286, located in the protein center below the P2 443 domain, can be detected (Figure 3 A). In contrast to GII.4 MI001 and GII.17 Kawasaki, GII.10 444 Vietnam P dimers show protection in the P2 domain including the canonical binding site (G451, 445 Y452) and the β-sheet region in the binding cleft, but also in the lower part of the P1 domain 446 (Figure 3 D). All observed differences can only be seen in the second, highly deuterated peak 447 distribution. The lowly deuterated peak distribution showed no significant differences between the 448 unbound and the glycan-bound state in any of the strains indicating that either only the highly 449 deuterated species can bind glycans or labeling time was too short to detect deuteration 450 differences in already strongly protected regions. 451 . Depicted are protein regions with significant differences in deuterium uptake between unbound P dimers and P dimers with either 10 mM HBGA B trisaccharide or 100 mM fucose (p < 0.05,

457
Student's T-test and ΔD > 2x pooled average SD). The deuteration difference in the second peak distribution 458 was manually validated by binomial fitting in case of bimodal spectra. Bar graphs and colored structures 459 indicate regions of P dimers, which get more protected (dark blue) or exposed (red) upon interaction with 460 glycans. Areas colored in grey showed no significant difference in the chosen HDX time regime and black 461 areas have no peptide coverage. P1/P2 refers to the two domains of the P dimer (shown in Figure 1).  binding is still possible in partially deamidated (iDN) or even fully deamidated (iDiD) P dimers at 469 the given concentration, even though binding is attenuated compared to the N373 wildtype [19]. 470

Influence of N373 deamidation on dynamics and glycan binding of GII.4
Occupation of the canonical binding sites has been seen in crystal structures of deamidated GII.4 471 Saga P dimers at elevated concentrations of 600 mM fucose, but binding interactions were slightly 472 different from wildtype [19]. In contrast to wild type GII.4 MI001 P dimer, no other region was 473 protected from HDX under fucose treatment, but increased deuteration in the main peak 474 distribution was observed in the P2 domain of both GII.4 strains, suggesting a more exposed 475 conformation. Interestingly, residues 335-362, which are protected in the wildtype proteins of all 476 strains, show increased deuteration in partially deamidated GII.4 P dimers. As we have a mixture 477 of wildtype and deamidated P domains in the sample, the mass shift we see in the deuterated 478 spectra reflects the average of all components, unbound and fucose-bound NN, iDN and iDiD 479 P dimers, which cannot be discriminated. However, binding probability calculations can give a 480 hint, which species contribute most to the observed increase in deuteration in presence of fucose. In GII.4 MI001 regions with bimodal peak distributions ( Figure 4D), relative intensities of both 489 distributions are similar to the ones in the wildtype protein. However, interaction with fucose in the 490 partially deamidated GII.4 MI001 P dimer does not lead to a significant increase in the relative 491 intensity of the lower deuterated peak distribution, as seen in the wildtype protein ( Figure 4D). 492 Slight bimodality is also present in peptides covering the canonical fucose binding site in the 493 fucose bound state but relative intensities are similar to the ones observed for the wildtype protein. P dimers (pdb 6H9V). In contrast to the wildtype N373 P dimer, parts of the P2 domain get more exposed 510 upon interaction with 100 mM fucose. (C) HDX differences in fully deamidated GII.4 Saga P dimers in 511 presence of 10 mM HBGA B trisaccharide are shown for comparison [19]. (D) Bimodality occurs in similar 512 regions as for the wildtype P dimer, implying that the more protected population is also present in the partially 513 deamidated sample. In contrast to the native P dimer, the relative intensity of the first peak distribution does 514 not significantly increase under fucose treatment (for statistics refer to description of Figure 3). (E) The 515 partially deamidated GII.4 MI001 N373iD P dimer shows a higher deuterium uptake in large parts of the 516 structure, which points towards higher flexibility, like in the fully deamidated GII.4 Saga P dimer [19] (F).

MD simulations
518 MD simulations were utilized to further investigate the norovirus P dimer strains. The RMSD 519 relative to the starting structure post-equilibration were calculated in order to estimate protein 520 dynamics during MD simulations. RMSF calculations were employed to examine fluctuations 521 throughout the simulated time frame, and highlight alterations in flexible regions of the different 522 protein chains. As support for the RMSF data, we calculated the Asas of the P dimers during the 523 simulation with respect to their crystal structure, providing an understanding of an increase or 524 decrease of the surface area of each individual residue. 525 The RMSD for the four norovirus P dimer strains are reported in Figure S18, in which the simulated 526 GII.4 Saga and MI001 P dimers reached a value of 1.5 Å after 100 ns. GII.10 Vietnam P dimers 527 show a maximum deviation around 90 ns at 2 Å, decreasing to 1.8 Å after 100 ns. GII.17 Kawasaki 528 P dimers demonstrate a still slightly increasing trend at the end of the simulation, indicating that 529 this system has not yet fully adapted to the solution environment. The RMSFs and Asas relative to 530 the respective crystal structures reveal differences in protein chain flexibility of the norovirus 531 strains, as depicted in Figure 5 and S17. The sequences were aligned for a better comparison. 532 Hence, resulting gaps in the individual RMSF graphs are due to missing residues at that specific 533 position. Least stability is introduced for the GII.4 Saga strain, as the RMSF values suggest only 534 a limited increase in fluctuation during the 100 ns of simulation. GII.4 MI001 and GII.17 Kawasaki 535 follow a similar trend. In contrast, GII.10 Vietnam P dimers show overall higher flexibility compared 536 to the other strains, in particular a peak around residue 350 in the P2 domain. Similar trends can 537 be observed for the Asas graphs depicted in Figure S17. GII.10 Vietnam demonstrates the highest 538 area values, which support the peaks observed in the RMSF graph in Figure 5A. The various P 539 dimer structures were overlaid in the PyMOL software, where areas of interest were imaged in 540 order to further explore differences of the protein crystal structures ( Figure S19). 541 Investigating GII.4 Saga P dimers complexed by fucose, and the potential role of deamidation, the 542 data revealed minimal difference between the RMSD values of said systems. RMSDs for the NN, 543 iDN and iDiD G.II 4 Saga dimers show a similar trend, reaching a value between 1.6 and 1.75 Å 544 after 100 ns of simulation ( Figure S20). RMSF and Asas calculations to investigate the role of 545 deamidation in the GII.4 Saga strains show that the individual graphs follow a similar trend, 546 suggesting only limited influence of the deamidation on the overall P dimer structure ( Figure S21-547 23) which is in line with previous crystallography data [19].

557
In this study, we address the differences in structural responses to glycan binding of norovirus P 558 dimers of the Asian epidemic strain GII.17 Kawasaki 308, the rarely detected strain GII.10 Vietnam 559 026 and the GII.4 MI001 strain, which belongs to the highly pandemic GII.4 genotype and has 560 been shown to infect mice as well [21]. 561 Bimodal peak distributions could originate from P particle formation 562 In our glycan-binding data, we observe the presence of bimodal peak distributions in a large 563 variety of peptides located in the P2 domain and the lower part of the P1 domain in all strains but 564 the previously analyzed GII.4 Saga. The intensity of the lower deuterated peak distribution was 565 between 7% and 17 % and relative intensities of both distributions remained constant over time. 566 This observation points towards two distinct protein populations [26] that experience a different 567 level of HDX over the whole exchange period. The low deuteration of the first peak distribution 568 suggests the presence of a compact conformation that is shielded from HDX. This could be in 569 principle true for protein aggregates, however, a protection of only the top and bottom part of the 570 protein appears very distinct. In case of protein aggregation, we would expect bimodality all over 571 the protein surface. 572 So, if the lowly deuterated population is no artifact, what could it be instead? The P domain can 573 form larger oligomers of different stoichiometry, up to whole 24-mer P particles, depending on the 574 protein concentration [53,54]. P oligomers form contacts through interactions in the lower part of 575 the P1 domain of each P dimer [53], which could explain the reduced deuteration in this area. 576 Closer inspection of the cryo EM structure [53] also suggests more contacts between the P2 577 domains compared to free P dimer. Importantly, the absence of bimodality in GII.4 Saga P dimers 578 implies that this strain has a different ability to form P oligomers than the closely related GII.4 579 MI001 strain. 580 P particles can bind HBGAs and are even suspected to interact with them in the same way as 581 VLPs [53,55]. However, in the presence of glycans no significant deuteration difference could be 582 detected in low deuterated peak distributions. Nevertheless, this does not explicitly mean that 583 there is no binding in these areas. Interpretation of no significant deuteration difference on a 584 certain time scale as the absence of any structural change should be treated with caution for 585 several reasons [28]. First, the intensity and thus the signal-to-noise ratio in the lower deuterated 586 peak distribution is low, which makes it difficult to detect statistically significant changes in 587 deuteration. Secondly, HDX is reduced in the lower deuterated population compared to P dimers 588 and would therefore need longer exchange times to reflect potential differences. In addition, peak 589 distributions in peptides covering the canonical glycan binding site are unimodal and show 590 protection in all strains, meaning this interaction can be found in the entire protein population. 591 In our data there is a clear increase of the potential P particle population in presence of 100 mM 592 fucose, which could mean that interaction with glycans supports the formation of P particles that 593 is otherwise less pronounced [55]. It has to be noted that we did not observe P particle oligomers 594 in native MS of a 4.5 µM P dimer solution and our protein constructs lack the C-terminal arginine 595 cluster that has been shown to be important for P particle formation [56,57]. However, the < 20% 596 of monomers, assumingly assembled into 24-mer P particles in absence of glycans, would amount 597 to around 1% of total signal intensity split up into many charge states in native MS, which likely 598 drop below detection limit. In contrast, fractions of structural variants of less than 5 % can be 599 detected in a properly conducted HDX-MS experiment [58]. 600

602
Apart from the presence of two distinct protein species in the sample, we could also detect 603 differences in glycan-induced protein dynamics in different strains. For GII.4 MI001, protected 604 regions were almost identical to the earlier investigated GII.4 Saga P dimer (canonical binding site 605 G443, Y444 and residues 283-303) [19], apart from additional protection in the upper P2 binding 606 cleft (residues 333-353). Involvement of this region has been seen in NMR data of GII.4 Saga P 607 dimers as well [19]. Furthermore, a recent NMR study suggests identical glycan binding behavior 608 of both GII.4 strains [49]. The same study also shows that MNV P dimers do not bind HBGAs, 609 underscoring that infectivity of GII.4 MI001 in mice cannot be explained by different glycan-induced 610 dynamics between GII.4 Saga and MI001 in line with our observations. 611 GII.17 Kawasaki P dimer crystal structures with fucose and HBGA A trisaccharide show backbone 612 interactions in T348 and G443 and side chain interactions in R349, D378 and Y444 [11, 59]. When 613 incubated with HBGA B trisaccharide and fucose, protection from HDX is observed for residues 614 333-353 corresponding to interactions with T348 and R349. In the presence of 100 mM fucose, 615 the canonical binding site (G443, Y444) is protected, as well as residues 269-286, which cannot 616 be explained by the known interactions from the crystal structures. This region is located below 617 the glycan binding cleft in the protein center, so protection from HDX could rather be the result of 618 a long-distance structural change than of direct interaction with fucose. It would be interesting to 619 see how long-distance structural changes would further propagate into the S domain in VLPs and 620 if they would influence the dynamic P domain lift off from the S domain that has been seen for 621 different norovirus strains [60][61][62]. 622 For the GII.10 Vietnam strain, binding of two HBGA B trisaccharide molecules and up to four 623 fucose molecules has been seen in crystal structures [10,12]. Compared to GII.4 MI001 and 624 GII.17 Kawasaki, we see protection in more protein areas for both HBGA B trisaccharide and 625 fucose, which mainly corresponds to the known glycan interactions summarized in Table 1. Due to close proximity of interacting amino acids in fucose binding sites 1/2 and 3/4 we cannot 627 distinguish these binding sites in HDX data at peptide resolution, but occupation of all four binding 628 sites is likely at the given concentration [12]. Interestingly, we see a protection of several residue 629 stretches that cannot be explained by known glycan interactions.  compared to the other strains that leads to smaller deuteration changes that are below the 655 detection limit in the current experimental setup. We also noticed that the GII.17 Kawasaki 656 datasets have a higher back exchange (D/H) than the other datasets so that small glycan induced 657 deuteration changes are more likely to be lost during the measurement. GII.4 and GII.10 P dimers 658 show protection of residues 285-298, which is also absent in GII.17 P dimers. Interestingly, P 659 dimers of the more prevalent strains GII.4 and GII.17 [3,4] show less changes in HDX upon glycan 660 binding compared to GII.10 Vietnam, which is rarely detected in patients [10]. 661 The RMSD plots for the four investigated P dimer strains without ligand reveal minimal differences 662 between all systems ( Figure S22). Whilst GII.4 Saga and MI001 trends reach a plateau, one can 663 observe a still increasing trend GII.17 Kawasaki. This indicates that this system has not yet 664 reached a stable conformation. The GII.10 Vietnam shows a decrease towards the end of the simulation, indicating that this system just adapted to the environment and obtained a stable 666

structure. 667
For the MD simulations, we were interested in the dynamics in absence of fucose, and we 668 observed an increase in flexibility throughout the peptide chain, accompanied with changes of the 669 Asas, which we present in Figure 5. The most prominent difference between the strains is a high 670 peak around the glycan binding site near residue 350 exclusively in GII.10 Vietnam P dimers. 671 GII.10 Vietnam has a longer loop around residue 350, which could explain the higher flexibility in 672 absence of fucose ( Figure 5). On further investigation however, when complexed by the ligand in 673 the crystal structure, this loop adopts a short helical structure ( Figure 6A), forming a pocket 674 shielding nearby residues from deuterium exchange, as observed in the HDX-MS experiment 675 ( Figure 3B). The crystal structures of the other strains have more unstructured loops. In our 676 simulations of ligand-free GII.10 Vietnam P dimers, the loop becomes flexible and unstructured, 677 as seen by the high RMSF values of up to 7.5 Å, and evident from snapshots taken from the MD 678 trajectory ( Figure 6B), which would explain the protection provided by bound ligands. This is 679 further supported by the increase of area accessible by the solvent ( Figure 5B). As such, for GII.10

696
The role of N373 deamidation 697 For GII.4 Saga P dimers it was observed earlier that spontaneous transformation of N373 into iso-698 aspartate (iD) attenuates glycan binding in fully deamidated iDiD P dimers. Deamidation is site-699 specific and happens over a timescale of 1-2 days at pH 7.3 and 37°C correlating with the length 700 of the infection cycle [19]. The high specificity of this deamidation under infection conditions 701 suggested that this process could occur as well in vivo, however, the biological relevance for 702 infection remained unclear. 703 To test if this effect can also be found in the closely related GII.4 MI001 strain, we performed HDX-704 MS on a spontaneously deamidated P dimer sample, resulting in mixed populations of NN, iDN 705 and iDiD dimers, which is more likely to be found in a natural infection context. For comparison, 706 we also measured fucose binding to a partially deamidated GII.4 Saga P dimer, which contained 707 an even higher fraction of fully deamidated iDiD P dimers. Strikingly, protection of the canonical 708 glycan binding site could still be detected in both isolates under fucose treatment, mainly 709 corresponding to binding to iDN and iDiD P dimers. Additionally, the P2 cleft was more exposed 710 in the partially deamidated sample under fucose treatment, in contrast to the protection observed 711 in the wildtype NN P dimer. This indicates that under natural deamidation conditions, glycan 712 binding at the canonical binding site still happens, but induces different dynamics than in the purely 713 wildtype P dimer. The exposure of the P2 cleft suggests that after glycan binding this area gets 714 more flexible, which could be required to interact with other factors or the until now unknown 715 receptor. As such an increase in deuteration is not present in wildtype NN P dimers, this effect 716 must be caused either by direct binding to deamidated P domains in iDN or iDiD dimers or by 717 binding to wildtype P domains in iDN dimers, whose overall dynamics are altered by the influence 718 of the neighboring deamidated monomer. RMSD, RMSF and Asas calculation for the GII.4 NN, iDN 719 and iDiD Saga P dimers show no striking differences when compared to each other and follow a 720 similar trend ( Figure S20-23). The fact that our MD simulations were unable to detect differences 721 between deamidated and non-deamidated P dimers suggests that any differences in dynamics 722 are manifested on timescales longer than a few 100 ns. Neither protection of binding sites nor 723 increased deuteration in P2 has been seen in previous HDX-MS measurements of fully 724 deamidated GII.4 Saga P dimer with 10 mM HBGA B trisaccharide under nearly identical 725 conditions [19]. A possible explanation could be that HBGA B trisaccharide concentration was too low to induce the observed effects because of decreased binding affinity. Notably, NMR 727 measurements of fully deamidated GII.4 Saga P dimers with HBGA B trisaccharide and fucose 728 show large chemical shift perturbations around residues 370-380 [19], a region where we observe 729 increased deuteration in presence of fucose. 730 We hypothesize that N373 deamidation serves as a pH and temperature dependent mechanism 731 to control infectivity of the virus. P dimers and VLPs have been shown to be stable under low pH 732 conditions and temperature [63], where deamidation rate is low [19]. After entering the human 733 host via contaminated food and reaching the intestine, the rise in pH and temperature facilitates 734 conversion of pure wildtype to partially deamidated P dimers that are still able to attach to glycans 735 and perform the structural change potentially required for interaction with the receptor and 736 infection of the target cell. This theory is supported by the observation of increased flexibility in the 737 P2 cleft of iDiD GII.4 Saga P dimers compared to the wildtype [19], which is also present in GII.4 738 MI001 P dimers. In summary, this could mean that deamidation creates the required flexibility for 739 host cell attachment and subsequent receptor binding. Attenuation of glycan binding in the 740 deamidated P dimer could be counteracted by avidity due to high glycan presentation on cell 741 surfaces in vivo. Native MS measurements of deamidated GII.4 Saga and MI001 P dimers also 742 show that with increasing deamidation, dissociation into monomers occurs, whereas in NN P 743 dimers no monomers are present ( Figure S10). This could as well be linked to the increased 744 flexibility of iDiD P dimers that weaken the dimer interface and shift the monomer-dimer 745 equilibrium. It would be interesting to investigate whether increased flexibility is limited to the P 746 domain or whether it is propagated into the S domain in VLPs as well, which as a result could 747 destabilize the particle and prepare for uncoating. 748 The question remains, which advantage the evolutionary conserved N373 deamidation site 749 provides for the most prevalent GII.4 strains over other strains. One possibility is that higher 750 flexibility induced by deamidation indeed enables better interactions with host receptors; another 751 possibility is that it is part of an immune escape mechanism. N373 is located in the