The oligomannose N-glycans 3D architecture and its response to the FcγRIIIa structural landscape

Oligomannoses are evolutionarily the oldest class of N-glycans, where the arms of the common pentasaccharide unit, i.e. Man α1-6)-[Manα(1-3)]-Manβ(1-4)-GlcNAcβ(1-4)-GlcNAcβ1-Asn, are functionalized exclusively with branched arrangements of mannose (Man) monosaccharide units. In mammalian species oligomannose N-glycans can have up to 9 Man, meanwhile structures can grow to over 200 units in yeast mannan. The highly dynamic nature, branching complexity and 3D structure of oligomannoses have been recently highlighted for their roles in immune escape and infectivity of enveloped viruses, such as HIV-1 and SARS-CoV2. The architectural features that allow these N-glycans to perform their functions is yet unclear, due to their intrinsically disordered nature that hinders their structural characterization. In this work we will discuss the results of over 54 µs of cumulative sampling by molecular dynamics (MD) simulations of differently processed, free (not protein-linked) oligomannose N-glycans common in vertebrates. We then discuss the effects of a protein surface on their structural equilibria based on over 4 µs cumulative MD sampling of the fully glycosylated CD16a Fc gamma receptor (FcγRIIIa), where the type of glycosylation is known to modulate its binding affinity for IgG1s, regulating the antibody-dependent cellular cytotoxicity (ADCC). Our results show that the protein’s structural constraints shift the oligomannoses conformational ensemble to promote conformers that satisfy the steric requirements and hydrogen bonding networks demanded by the protein’s surface landscape. More importantly, we find that the protein does not actively distort the N-glycans into structures not populated in the unlinked forms in solution. Ultimately, the highly populated conformations of the Man5 linked glycans support experimental evidence of high levels of hybrid complex forms at N45 and show a specific presentation of the arms at N162, which may be involved in mediating binding affinity to the IgG1 Fc.


Introduction
Complex carbohydrates (or glycans) are the most abundant biomolecules in nature. Within a human biology context, glycans coat cell membranes and protein surfaces, mediating a myriad of essential biological processes in health and disease states [1][2][3][4][5][6] . N-glycosylation is one of the most abundant and diverse type of post-translational modification that can affect protein trafficking, structural stability and mediate interactions with different receptors [6][7][8][9][10][11] . N-glycan recognition and binding affinities are often highly specific to their sequence, intended as the types of monosaccharides, their stereochemistry and branching patterns 12 ; a principle that has been successfully exploited in the development of glycan microarray technology 13 .
Molecular recognition is fundamentally dependent, among other considerations, on structural and electrostatic complementarity between the ligand and the receptor's binding site. Within this framework, the prediction and characterization of glycan binding specificity is an extremely difficult task, due to their high degree of flexibility or intrinsic disorder, which hinders our ability to determine their 3D structure by means of experimental techniques.
Indeed, glycans can only be structurally resolved in their entirety only when tightly bound to a receptor, thus when their conformational degrees of freedom are heavily restrained. Because of their inherent flexibility, free glycans can adopt different 3D structures within a weighted conformational ensemble, which cannot be determined with currently available experimental methods; although very promising steps forward have been recently made in advancing imaging techniques for single glycans 14,15 .
High performance computing (HPC) molecular simulations can contribute a great deal towards our understanding of the relationships between glycans' sequence, structure and function.
Indeed, conformational sampling through conventional and/or enhanced molecular dynamics (MD) schemes allows us to characterize the dynamic behaviour of different glycoforms at the atomistic level of details. Within this context, for the past few years our lab contributed to the knowledge of N-glycans dynamics by providing information on their 3D architecture and relative flexibility from extensive MD-based conformational sampling 16,17 . As an example, we have shown how the sequence (and branching) of complex N-glycans determines the 3D structure, which in turn drives their recognition 16,17 . In this work we extend our dataset of free (unlinked) N-glycans structures to the vertebrate oligomannose type, where, as shown in units. In addition, we also determine how the protein surface landscape affects their conformational dynamics, which is a very important question in terms of its impact on molecular recognition and function, while challenging to answer in absolute terms because of the site-specific character. Oligomannoses are often defined as "immature" N-glycans, as they are processed towards complex functionalization in the Golgi 6 and are not abundant in vertebrates. Nevertheless, these N-glycans are a common post-translational modification of viral envelope proteins expressed in human cell lines 20,21 , for example it is the prevalent type of glycosylation of the HIV-1 fusion trimer [22][23][24][25] . Furthermore, an increase in large oligomannose-type N-glycosylation in humans has been linked to breast cancer progression [26][27][28] and can occur where the protein landscape at the N-glycan site does not allow easy access to the required glycohydrolases and glycotransferases for further functionalization 6,29,30 . Interestingly, recent work has shown that oligomannose N-glycans functionalizing CD16a low-affinity Fc g receptors (FcgIIIa) determine an increase in IgG1-binding affinity by 51-fold 31 , relative to the more common complex N-glycans 32 , although the N-glycosylation composition varies depending on the glycosylation site 32 .
In this work we have studied the effect of the FcgIIIa protein surface landscape on the intrinsic conformational propensity of different oligomannose N-glycans we determined for the unlinked forms. Our results show that the two FcgIIIa N-glycosylation sites, N45 and N162, affect the oligomannose dynamics rather differently, in function of the structural constrains of the sites and of the 3D architecture of the glycan. More specifically, we find that the protein landscape affects the glycans conformational equilibrium by promoting structures that are complementary to it and not by actively changing their intrinsic architecture. Indeed, all the 3D conformers observed in the analysis of the bound oligomannoses, are always identified in the simulations of the corresponding unlinked forms in solution, although in different populations.
Interestingly, we also determined that the progressive elongation of the arms/branches promotes inter-arm contacts, where the Man9 3D architecture is almost entirely structured with interacting arms. Finally, these findings fit very well within the framework of our recently proposed "glycoblocks" glycans structure representation 16 , whereby groups of specifically linked monosaccharides within N-glycans represent independent structural elements (or glycoblocks), which exposure, or presentation in function of the particular protein landscape, drives molecular recognition.  36,37 , and also because of consistency with our previous work 16,17 , we consider the choice of GLYCAM06-j1/TIP3P parameter set as appropriate. All simulations were run in 200 mM NaCl salt concentration, with counterions represented by amber parameters 38 in a cubic simulation box of 16 Å sides. Long range electrostatic were treated by Particle Mesh Ewald (PME) with cut-off set at 11 Å and a B-spline interpolation for mapping particles to and from the mesh of order of 4. Van der Waals (vdW) interactions were cut-off at 11 Å. The MD trajectories were generated by Langevin dynamics with collision frequency of 1.0 ps -1 . Pressure was kept constant by isotropic pressure scaling with a pressure relaxation time of 2.0 ps. After an initial 500.000 cycles of steepest descent energy minimization, with all protein/glycans heavy atoms restrained by a harmonic potential with a force constant of 5 kcal mol -1 Å -2 , the system was heated in two stages, i.e. from 0 to 100 K over 500 ps at constant volume and then from 100 K to 300 K over 500 ps at constant pressure. After the heating phase all restraints were removed and the system was allowed to equilibrate for 5 ns at 300 K and at 1 atm of pressure. Production and subsequent analysis was done on 500 ns trajectories run in parallel for each uncorrelated starting structure, i.e. each conformer generated with GLYCAM-Web. Analysis was done using the cpptraj tool and with VMD 39  Figure 1c, and to the N162 sidechain, to obtain two systems, one with only Man5 and the other with only Man9 at both positions. As a note, the structure of the Nglycan at N165 from the crystal structure is quite distorted with uncommon conformations of some of the monosaccharides, probably determined by the fitting to the electron density, therefore it was disregarded and only the chitobiose was used for structural alignment. These systems were run in duplicates from uncorrelated starting structures with the same simulation protocol used for the free glycans. Production runs were extended to 1 µs for each trajectory, for a total of 4 µs of cumulative sampling time. All simulations were run on NVIDIA Tesla V100 16GB PCIe (Volta architecture) GPUs on resources from the Irish Centre for High-End Computing (ICHEC) (www.ichec.ie).

Results
We used conventional MD simulations, run in parallel for 500 ns from nine uncorrelated starting points 16,17 , to characterize the 3D structure and dynamics of human oligomannose Nglycans, when unlinked, see
As found for complex biantennary N-glycans 16 vice versa. We also identified two alternative, less populated conformers, namely a "front fold" (phi = 79°, psi = 87°) with a relative population of 12% and a "back fold" (phi = 83°, psi = -76°) with a relative population of 6%, see Figure 2 and Table Table   1, and does not show a back fold orientation.   Table 2 and Tables S.9 and S.13, with a slightly more pronounced preference for the back vs front fold in Man9, due to the interactions of the longer (1-6) branch with the chitobiose, see Figure 5 and Table 2.  Back fold 70 (7) -118 (12) -67 (10) 12
Man9 has two Mana(1-2)-Man linkages elongating both branches on the (1-6) arm, denying the pose found for Man5, which indeed disappears, see Tables 3 and S.13. Despite a higher flexibility relative to Man5, the N45-linked Man9 is less dynamic relative to the unlinked form due to the protein's landscape. Indeed, as shown in Table 3, only three out of the seven populated conformers are accessible.
Consequently, the intrinsic dynamics of the N162-Man5 is almost entirely retained, with a shift promoting the open (cluster 2) relative to the open (cluster 1) as the dominant conformer, see Table 4. Meanwhile in case of a N162-linked Man9, the dynamics of the longer arms is limited due to the proximity to the protein's surface, see Figure 8,  Man9 are shown on the right-hand side, with the protein represented by the solvent accessible surface and underlying cartoons in grey and the mannose residues with different shades of green as described in the legend. Heat maps made with RStudio (www.rstudio.com) and structure rendered with pyMol (www.pymol.org). N-glycan coloured according to the SNFG convention.

Discussion
In this work we analysed the 3D structure and dynamics of human oligomannose N-glycans, The results obtained for the unlinked oligomannoses also confirm an earlier observation we made in the context of complex N-glycans 16 , whereby the overall 3D architecture is determined by the local spatial arrangement of independent groups of monosaccharides, we named "glycoblocks". The oligomannoses dynamics can be also discretized in terms of these structural units 16 , with the addition of a unique Mana(1-2)-Man glycoblock that can be added to the arms, with as we have seen, minimal effect to the dynamics of the underlying units it builds on. This observation can offer a practical advantage to the study of glycan recognition through molecular docking for example, where the receptor binds a specific glycoblock unit and recognition depends only on its accessibility within a specific glycoform.
To understand how the protein affects the presentation of the glycans to potential receptors we have looked at the human FcgRIIIa (CD16a). Human FcgRIIIa has two N-glycosylation sites, namely N45 and N162, where the type of glycosylation affects the receptor's binding affinity to IgG1s 31,32,43 . The surface landscape around these two sites is quite different, with N162 exposed to the solvent, while N45 located in the core of one of the two structural domains, see  Table 4 and Figure 8. Ultimately, the comparison between the conformational propensity of the unlinked Man5 and Man9 oligomannoses relative to their FcgRIIIa-linked counterparts suggests that the protein landscape affects the glycans structure by shifting their intrinsic conformational equilibria towards forms that complement it, yet it does not actively morph the glycan into un-natural conformers.

Conclusions
In this work we have characterized the 3D structure and dynamics of human oligomannose N- which has been found to have an unusually high degree of hybrid N-glycoforms, and ultimately exposure of the arms at N162 for contact with the IgG1 Fc N-glycans, which is implicated in modulating ADCC 19,31,44,45 . Work in this direction is currently ongoing in our lab.

Supporting Information
The accompanying supporting information includes 17 Tables and 18 Figures providing a complete list of all rotamers conformations with relative populations for all oligomannoses isomers studied in this work. Additional data analysis is shown in Figure S.18.