Water Channel

The clinical laboratories (Innsbruck, Mayo Clinic, Oxford, and Sydney; centers 1–4) sent the following groups of coded serum samples and clinical information to the Institute for Quality Assurance (IfQ; Lübeck, Germany):

Phase I: 89 coded samples sent to centers 1–4 and center 5 (Euroimmun) for testing (figure 1)

MOG-IgG clearly positive: 39 blinded samples from all laboratories with a previously determined clearly positive MOG-Ab serostatus (high titers or fluorescence-activated cell sorting [FACS] binding ratios, supplementary methods, table e-2, links.lww.com/NXI/A189), all of them diagnosed with inflammatory demyelinating diseases known to be associated with MOG-IgG (such as ADEM, aquaporin-4 [AQP4] antibody–negative neuromyelitis optica spectrum disorder (NMOSD), optic neuritis, myelitis, and other demyelinating diseases).

MOG-IgG clearly negative (negative or very low titers or FACS binding ratios, supplementary methods, table e-2, links.lww.com/NXI/A189): 40 blinded samples from all laboratories with a previously determined clearly negative MOG-Ab serostatus. Eighteen of the 40 samples were from people who also presented with clinically overlapping features such as optic neuritis, myelitis, ADEM, or encephalitis. The other samples were from controls (7 from people with MS, 5 from people with other neurologic diseases, and 10 from healthy controls).

Ten technical controls (humanized monoclonal MOG-Ab 8-18-C5,^{30 (link)} 5 samples IgG1, and 5 samples IgM (kappa) in different dilutions, but of unknown IgG or IgM concentration, contributed by center 5.

Phase II: 100 coded samples sent to 5 centers for testing (18 repeat and 82 new, figure 1)

Nine positive and 9 negative samples from phase I were sent out a second time to assess interassay variations.

Thirty healthy blood donors were contributed by the IfQ. No clinical information was available, and samples were not pretested for antibodies against MOG or other autoantigens.

MOG-IgG low/borderline positive: 39 blinded samples from all laboratories with a previously determined low positive MOG-IgG serostatus (just above the individual cutoff values, supplementary methods, table e-2, links.lww.com/NXI/A189). Thirty-six of these samples were from people with inflammatory demyelinating diseases associated with MOG-IgG and 3 were from patients with MS.

MOG-IgG borderline negative: 13 blinded samples from all laboratories with a previously determined borderline negative MOG-IgG serostatus (just below the individual cutoff values, supplementary methods, table e-2, links.lww.com/NXI/A189). Five of these samples were from patients with inflammatory demyelinating diseases associated with MOG-IgG and 8 were from controls (3 from people with MS and 5 from people with other neurologic diseases).

BvAQPs were identified by HMM (Hidden Markov Model) and BLAST homology search. The sugar beet predicted proteome was collected using the sugar beet genome RefBeet-1.2 (http://bvseq.molgen.mpg.de/Genome/Download/index.shtml). The Hidden Markov model (HMM) of the MIP domain (PF00230) was downloaded from the Sanger database (http://pfam.xfam.org/family/PF00230). PF00230 was then used to query the predicted Sugar beet proteome using HMMER 3.0 software (http://hmmer.org/). 35 Arabidopsis AtAQPs were download from TAIR Database (https://www.arabidopsis.org/browse/genefamily/Aquaporins.jsp) and then used to search BvAQPs with BLASTp tool using NCBI sugar beet genome (https://www.ncbi.nlm.nih.gov/genome/?term=Beta+vulgaris) and sugar beet genome (http://www.genomforschung.uni-bielefeld.de/en/projects/annobeet) (Baranwal, Negi & Khurana, 2016 (link)) with cut-off E-value of e⁻⁵. All no-redundant gene sequences were analyzed by SMART (http://smart.embl-heidelberg.de/) and Pfam (http://pfam.xfam.org/search/sequence). Sequences encoding complete MIP domain and two NPA motifs were considered as putative AQP genes.
Additionally, the molecular weight (MW) and isoelectric point (PI) of BvAQPs were calculated by ExPASy (http://www.expasy.org/); transmembrane helical domains (TMHs) were assessed by TMHMM Sever v.2.0 (http://www.cbs.dtu.dk/services/TMHMM/); the subcellular localization of BvAQPs were predicted using Plant-mPLoc (http://www.csbio.sjtu.edu.cn/bioinf/plant-multi/) and WolF PSORT (http://www.genscript.com/wolf-psort.html). The position of the AQP genes on the sugar beet chromosomes were identified based on position information from the sugar beet genome database and the distribution graph of AQP genes was drawn by MapInspect software (http://mapinspect.software.informer.com/).

Schrodinger software was used to prepare the Panx2 model, including building the missing loop (amino acids 185–187, 267–272, and 374–376) via the Prime ^{60 (link)}, adding protons to all amino acids via PROPKA^{61 (link)}, and energy-minimizing the system with the OPLS4 force field^{62 (link)} to ensure that no positional conflicts were in the model. Subsequently, a web-based platform for generating the inputs for molecular dynamics named CHARMM-GUI^{63 (link)} was applied to build the system. The N- and C-termini of the model were treated as amino termini with a positive charge and carboxyl termini with a negative charge, respectively, as they were freely exposed to the solvation. Two intramolecular disulfide bonds (C81–C279, C99–C259) identified in the experimental structures were created for each chain. The state of the protein after the above processes was regarded as its initial geometry in our simulations. The orientation of the protein and the position of lipid bilayers were determined by the Positioning of Proteins in Membranes (PPM) 2.0 server^{64 (link)} and checked manually. The lipid bilayers with the 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) model and water molecules with the TIP3P model were generated, resulting in a 120 × 120 × 165 Å box with 200 POPC molecules. The pore water molecules (water in the ion channel) were also added based on the channel geometry. Finally, the entire system was neutralized with 150 mM NaCl.
GPU-accelerated Gromacs 2021^{65 (link)} was used to perform the molecular dynamics simulation and CHARMM36m force field^{66 (link)} for the molecules. The MD simulation consisted of energy minimization, pre-equilibration and production simulations. The system was first energy minimized with the steepest descent algorithm while keeping 4000 kJ/(mol nm²) force constant on backbone atoms and ligand atoms, 2000 kJ/(mol nm²) force constant on side-chain atoms and 1000 kJ/(mol nm²) force restraint on lipid atoms. Then, six-step pre-equilibration simulations (0.6 ns, 0.6 ns ps, 1 ns, 1 ns, 1 ns, and 1 ns) were carried out, where restraint was reduced slowly (4000, 2000, 1000, 500, 300, 0 kJ/(mol nm²) on the backbone and ligand atoms, 2000, 1000, 500, 200, 50, 0 kJ/(mol nm²) on side-chain atoms, and 1000, 400, 400, 200, 40, 0 kJ/(mol nm²) on lipid atoms) to relax the system. Finally, a production simulation was performed for 100 ns using Langevin thermostat, with a constant temperature of 310 K and a constant pressure of 1 atm. Periodic boundary conditions (PBCs) were introduced during the all molecular dynamics simulations. In all steps, the time step was 2 fs, and atomic coordinates were written every 5 ps. After the MD simulation, root mean squared deviations, root mean squared fluctuations and distances between atoms were analyzed by Gromacs. Details of MD simulations have also been deposited in github at [https://github.com/shiyu-wangbyte/panx2-simulation].