Fucose

L. hyperborea, L. digitata, and A. nodosum were harvested in February 2016 (Quality Sea Veg Ltd., Co. Donegal, Ireland). Samples were cleaned from epitopes, oven-dried following industry practices (50 °C, 9 days), and milled to 1 mm particle size using a hammer mill (Christy and Norris, Chelmsford, UK). All the samples were vacuum-packed and stored at room temperature for further analyses. The dry matter of the dried and milled macroalgae was determined by oven-drying the samples at 105 °C for 16 h. The ash content was determined after ignition of a weighed sample in a muffle furnace at 550 °C for 6 h according to the AOAC.942.05 [38 ]. The N content was determined using the LECO FP 528 instrument (Leco Instruments UKLTD., Cheshire, UK), using the conversion factor 4.17 as described for brown macroalgae by Biancarosa et al. [39 (link)]. The ether extract was determined using Soxtec instrumentation (Tecator, Sweden) following the AOAC.920.39 [38 ], and the total soluble sugars were estimated following the phenol-sulfuric acid assay as described by Brummer and Cui [40 ]. The measurements of fucose and total glucans were performed following the methodology described in Section 3.5.

For the analyses, EPS underwent acid hydrolysis followed by derivatization. Lyophilized EPS (1 mg) were solubilized in 250 μL of HCl 4 M and incubated at 99 °C for 2 h under gentle shaking (300 rpm, Thermomixer Comfort; Eppendorf, Milan, Italy) for hydrolysis. 250 μL of NaOH 4 M were then added to the samples for neutralization.
Derivatization of monosaccharides was carried out slightly modifying a previously reported procedure [25 ]. In details, 120 μL of the hydrolysed solutions were mixed with 180 μL of NaOH 0.5 M. 200 μL of the resulting samples were mixed with 200 μL of PMP (0.5 M in methanol) and incubated at 70 °C for 1 h under gentle shaking (300 rpm). After cooling at room temperature, 200 μL of HCl 0.3 M and 300 μL of Tris buffer (1.5 M, pH 7.00) were subsequentially added for neutralization. The resulting mixtures were extracted 3 times with 500 μL of dichloromethane to remove the excess of PMP. Samples were aliquoted and stored at − 20 °C until analysis. Standard solutions (6.25 mM) of d-mannose, d-glucosamine, d-galactosamine, d-rhamnose, d-glucose, d-galactose,d-xylose, and d-fucose were derivatized following the same procedure.
The whole procedure was performed in triplicate for each sample.

The atomistic models of adalimumab and avelumab were built with the same chimeric homology modeling approach by the “Homology model” tool of MOE software²⁶ . The crystallographic structures of both adalimumab and avelumab Fab domains are available (PDB IDs: 3WD5 and 4NKI, respectively)^{29 (link),30 (link)} and were used to model them, while the only solved structure of a whole human IgG1 was used as a template (PDB ID: 1HZH)^{31 (link)} to build hinge and Fc portions. The structure of 1HZH that we chose is the one contained in the MOE library of crystalized antibodies, which has been processed by MOE experts with MD simulations. This choice derives from the need to use the same starting conformation for the antibodies to point out differences mediated by glycans. Before performing homology modeling, all the templates were prepared using the “Structure Preparation” tool of MOE, to correct any crystallographic issue, and processed by the “Protonate 3D” tool to assign the ionization states and add missing hydrogens. Both the 3WN5 and 4NKI structures present some missed residues. The last three C-terminal residues in LC of 3WD5 (Gly212, Glu213, and Cys214) were modeled on 1HZH.pdb with the enabled “override template” option. In 4NKI, the first (Gln1) and the last three (Glu214, Cys215 and Ser216) amino acids of the LC were missed. Gln1 was manually added and minimized via the “Protein Builder” tool by MOE, Glu214 and Cys215 were instead modelled on 1HZH.pdb to preserve their orientation with the enabled “override template” option. Ser216 was added once the chimeric model was built and minimized together with the adjacent disulfide bond, in order to preserve the stability of the aglycosylated structure.
The models were then glycosylated. As reported previously^{20 (link)}, G0 and G0F sugars were attached unit by unit to the conserved Asn297 of adalimumab (Asn301) by the MOE “Carbohydrate builder”. A final minimization step was carried out on entire glycosylated models until the RMS gradient reached 0.01 kcal/mol/Ų. For avelumab, G0F glycans, already present in the 1HZH template of the MOE library, were linked to the conserved Asn297 (Asn300) after structural superposition and energy minimized down to an RMS gradient of 0.01 kcal/mol/A². To build the G0 models, the fucose was deleted from the glycan chains on the G0F models, then the glycans and Asn297 were energy minimized again down to an RMS gradient of 0.1 kcal/mol/A².